├── .gitattributes
├── .gitignore
├── .idea
    ├── RNN-LM.iml
    ├── inspectionProfiles
    │   └── profiles_settings.xml
    ├── misc.xml
    ├── modules.xml
    └── vcs.xml
├── LICENSE
├── README.md
├── data
    ├── .ipynb_checkpoints
    │   └── Data-Analysis-checkpoint.ipynb
    ├── Data-Analysis.ipynb
    └── train_labels.csv
├── main.py
├── model.py
├── package
    ├── __pycache__
    │   ├── config.cpython-37.pyc
    │   ├── data_loader.cpython-37.pyc
    │   ├── definition.cpython-37.pyc
    │   ├── evaluator.cpython-37.pyc
    │   ├── loss.cpython-37.pyc
    │   ├── trainer.cpython-37.pyc
    │   └── utils.cpython-37.pyc
    ├── config.py
    ├── data_loader.py
    ├── definition.py
    ├── evaluator.py
    ├── loss.py
    ├── trainer.py
    └── utils.py
└── preprocess
    ├── .ipynb_checkpoints
        └── Preprocess-checkpoint.ipynb
    ├── KoNPron.ipynb
    ├── Preprocess.ipynb
    ├── konpron.py
    ├── special.csv
    └── train_labels.csv


/.gitattributes:
--------------------------------------------------------------------------------
1 | docs/* linguist-vendored
2 | sphinx/* linguist-vendored
3 | preprocess/* linguist-vendored
4 | data/* linguist-vendored
5 | 


--------------------------------------------------------------------------------
/.gitignore:
--------------------------------------------------------------------------------
1 | *.bin
2 | *.zip
3 | *.idea
4 | __pycache__/
5 | *.pyc
6 | .idea


--------------------------------------------------------------------------------
/.idea/RNN-LM.iml:
--------------------------------------------------------------------------------
 1 | <?xml version="1.0" encoding="UTF-8"?>
 2 | <module type="PYTHON_MODULE" version="4">
 3 |   <component name="NewModuleRootManager">
 4 |     <content url="file://$MODULE_DIR$" />
 5 |     <orderEntry type="inheritedJdk" />
 6 |     <orderEntry type="sourceFolder" forTests="false" />
 7 |   </component>
 8 |   <component name="PyDocumentationSettings">
 9 |     <option name="format" value="GOOGLE" />
10 |     <option name="myDocStringFormat" value="Google" />
11 |   </component>
12 |   <component name="TestRunnerService">
13 |     <option name="PROJECT_TEST_RUNNER" value="pytest" />
14 |   </component>
15 | </module>


--------------------------------------------------------------------------------
/.idea/inspectionProfiles/profiles_settings.xml:
--------------------------------------------------------------------------------
1 | <component name="InspectionProjectProfileManager">
2 |   <settings>
3 |     <option name="USE_PROJECT_PROFILE" value="false" />
4 |     <version value="1.0" />
5 |   </settings>
6 | </component>


--------------------------------------------------------------------------------
/.idea/misc.xml:
--------------------------------------------------------------------------------
1 | <?xml version="1.0" encoding="UTF-8"?>
2 | <project version="4">
3 |   <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.7 (Korean-Speech-Recognition) (2)" project-jdk-type="Python SDK" />
4 | </project>


--------------------------------------------------------------------------------
/.idea/modules.xml:
--------------------------------------------------------------------------------
1 | <?xml version="1.0" encoding="UTF-8"?>
2 | <project version="4">
3 |   <component name="ProjectModuleManager">
4 |     <modules>
5 |       <module fileurl="file://$PROJECT_DIR$/.idea/RNN-LM.iml" filepath="$PROJECT_DIR$/.idea/RNN-LM.iml" />
6 |     </modules>
7 |   </component>
8 | </project>


--------------------------------------------------------------------------------
/.idea/vcs.xml:
--------------------------------------------------------------------------------
1 | <?xml version="1.0" encoding="UTF-8"?>
2 | <project version="4">
3 |   <component name="VcsDirectoryMappings">
4 |     <mapping directory="$PROJECT_DIR$" vcs="Git" />
5 |   </component>
6 | </project>


--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
 1 | MIT License
 2 | 
 3 | Copyright (c) 2020 SooHwan Kim
 4 | 
 5 | Permission is hereby granted, free of charge, to any person obtaining a copy
 6 | of this software and associated documentation files (the "Software"), to deal
 7 | in the Software without restriction, including without limitation the rights
 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 9 | copies of the Software, and to permit persons to whom the Software is
10 | furnished to do so, subject to the following conditions:
11 | 
12 | The above copyright notice and this permission notice shall be included in all
13 | copies or substantial portions of the Software.
14 | 
15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 | SOFTWARE.
22 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # Character RNN Language Model
 2 |   
 3 | ### Recurrent Neural Netowrk Language Model in Pytorch  
 4 | [<img src="https://github.com/gentaiscool/end2end-asr-pytorch/raw/master/img/pytorch-logo-dark.png" height=18>](https://pytorch.org/) <img src="https://img.shields.io/badge/License-MIT-yellow" height=20>
 5 |   
 6 | ## Intro
 7 |   
 8 | This is a project for Character-level Recurrent Neural Network Language Model (Rnnlm) implemented in [Pytorch](pytorch.org).  
 9 | This language model can be combined with tasks such as speech recognition, machine translation, image captioning etc..   
10 | We appreciate any kind of feedback or contribution.
11 |   
12 | <img src="https://user-images.githubusercontent.com/42150335/79066605-39189c00-7cf4-11ea-8d92-971a360f3644.png" width=620>  
13 |   
14 | ## Installation
15 | This project recommends Python 3.7 or higher.   
16 | I recommend creating a new virtual environment for this project (using virtualenv or conda).  
17 | 
18 | ### Prerequisites
19 |   
20 | * Numpy: `pip install numpy` (Refer [here](https://github.com/numpy/numpy) for problem installing Numpy).
21 | * Pandas: `pip install pandas` (Refer [here](https://github.com/pandas-dev/pandas) for problem installing Pandas)
22 | * PyTorch: Refer to [PyTorch website](http://pytorch.org/) to install the version w.r.t. your environment.
23 |   
24 |   
25 | ## Troubleshoots and Contributing
26 |   
27 | If you have any questions, bug reports, and feature requests, please [open an issue](https://github.com/sooftware/RNN-LM/issues) on Github.  
28 | or Contacts sh951011@gmail.com please.  
29 | 
30 | I appreciate any kind of feedback or contribution. Feel free to proceed with small issues like bug fixes, documentation improvement. For major contributions and new features, please discuss with the collaborators in corresponding issues.
31 |   
32 | ### Code Style
33 |   
34 | We follow PEP-8 for code style. Especially the style of docstrings is important to generate documentation.  
35 |   
36 | ## Reference
37 |   
38 | [[1] IBM pytorch-seq2seq](https://github.com/IBM/pytorch-seq2seq)  
39 | [[2] Character-unit based End-to-End Korean Speech Recognition](https://github.com/sooftware/End-to-End-Korean-Speech-Recognition)  
40 | [[3] 「An analysis of incorporating an external language model into a sequence-to-sequence model」 Paper](https://arxiv.org/abs/1712.01996)  
41 |   
42 | ## Author
43 |   
44 | * Soohwan Kim [@sooftware](https://github.com/sooftware)
45 | * Contacts: sh951011@gmail.com
46 | 


--------------------------------------------------------------------------------
/data/.ipynb_checkpoints/Data-Analysis-checkpoint.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "metadata": {},
  7 |    "outputs": [],
  8 |    "source": [
  9 |     "import pickle\n",
 10 |     "import pandas as pd\n",
 11 |     "import matplotlib.pyplot as plt\n",
 12 |     "import seaborn as sns"
 13 |    ]
 14 |   },
 15 |   {
 16 |    "cell_type": "code",
 17 |    "execution_count": 2,
 18 |    "metadata": {},
 19 |    "outputs": [
 20 |     {
 21 |      "data": {
 22 |       "text/html": [
 23 |        "<div>\n",
 24 |        "<style scoped>\n",
 25 |        "    .dataframe tbody tr th:only-of-type {\n",
 26 |        "        vertical-align: middle;\n",
 27 |        "    }\n",
 28 |        "\n",
 29 |        "    .dataframe tbody tr th {\n",
 30 |        "        vertical-align: top;\n",
 31 |        "    }\n",
 32 |        "\n",
 33 |        "    .dataframe thead th {\n",
 34 |        "        text-align: right;\n",
 35 |        "    }\n",
 36 |        "</style>\n",
 37 |        "<table border=\"1\" class=\"dataframe\">\n",
 38 |        "  <thead>\n",
 39 |        "    <tr style=\"text-align: right;\">\n",
 40 |        "      <th></th>\n",
 41 |        "      <th>ko</th>\n",
 42 |        "      <th>id</th>\n",
 43 |        "    </tr>\n",
 44 |        "  </thead>\n",
 45 |        "  <tbody>\n",
 46 |        "    <tr>\n",
 47 |        "      <th>0</th>\n",
 48 |        "      <td>어디 보자...</td>\n",
 49 |        "      <td>8 190 0 42 45 1 1 1</td>\n",
 50 |        "    </tr>\n",
 51 |        "    <tr>\n",
 52 |        "      <th>1</th>\n",
 53 |        "      <td>칠대 왕국 종족 중에 오크는 처음 듣는다</td>\n",
 54 |        "      <td>318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...</td>\n",
 55 |        "    </tr>\n",
 56 |        "    <tr>\n",
 57 |        "      <th>2</th>\n",
 58 |        "      <td>이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.</td>\n",
 59 |        "      <td>3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...</td>\n",
 60 |        "    </tr>\n",
 61 |        "    <tr>\n",
 62 |        "      <th>3</th>\n",
 63 |        "      <td>라니스터 가의</td>\n",
 64 |        "      <td>32 20 79 162 0 6 130</td>\n",
 65 |        "    </tr>\n",
 66 |        "    <tr>\n",
 67 |        "      <th>4</th>\n",
 68 |        "      <td>별 희한한 생각이 다 떠오르곤 하죠</td>\n",
 69 |        "      <td>233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...</td>\n",
 70 |        "    </tr>\n",
 71 |        "    <tr>\n",
 72 |        "      <th>...</th>\n",
 73 |        "      <td>...</td>\n",
 74 |        "      <td>...</td>\n",
 75 |        "    </tr>\n",
 76 |        "    <tr>\n",
 77 |        "      <th>2992115</th>\n",
 78 |        "      <td>네 오늘 아침 범죄현장에 있었죠</td>\n",
 79 |        "      <td>96 0 57 275 0 5 373 0 560 700 350 109 17 0 26 ...</td>\n",
 80 |        "    </tr>\n",
 81 |        "    <tr>\n",
 82 |        "      <th>2992116</th>\n",
 83 |        "      <td>머시</td>\n",
 84 |        "      <td>235 47</td>\n",
 85 |        "    </tr>\n",
 86 |        "    <tr>\n",
 87 |        "      <th>2992117</th>\n",
 88 |        "      <td>눈은 풀려 있었고 입에선 연신 침이 흘러 나왔다.</td>\n",
 89 |        "      <td>351 23 0 449 108 0 26 62 7 0 219 17 194 0 147 ...</td>\n",
 90 |        "    </tr>\n",
 91 |        "    <tr>\n",
 92 |        "      <th>2992118</th>\n",
 93 |        "      <td>나는 좋은 선생님이야.</td>\n",
 94 |        "      <td>13 4 0 94 23 0 194 71 216 3 25 1</td>\n",
 95 |        "    </tr>\n",
 96 |        "    <tr>\n",
 97 |        "      <th>2992119</th>\n",
 98 |        "      <td>다만 취업 준비를 지원한다는 제도 성격을 고려해 유흥.도박.성인 용품 등 용도나 고...</td>\n",
 99 |        "      <td>15 46 0 331 207 0 245 122 55 0 10 82 27 15 4 0...</td>\n",
100 |        "    </tr>\n",
101 |        "  </tbody>\n",
102 |        "</table>\n",
103 |        "<p>2992120 rows × 2 columns</p>\n",
104 |        "</div>"
105 |       ],
106 |       "text/plain": [
107 |        "                                                        ko  \\\n",
108 |        "0                                                 어디 보자...   \n",
109 |        "1                                   칠대 왕국 종족 중에 오크는 처음 듣는다   \n",
110 |        "2               이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.   \n",
111 |        "3                                                  라니스터 가의   \n",
112 |        "4                                      별 희한한 생각이 다 떠오르곤 하죠   \n",
113 |        "...                                                    ...   \n",
114 |        "2992115                                  네 오늘 아침 범죄현장에 있었죠   \n",
115 |        "2992116                                                 머시   \n",
116 |        "2992117                        눈은 풀려 있었고 입에선 연신 침이 흘러 나왔다.   \n",
117 |        "2992118                                       나는 좋은 선생님이야.   \n",
118 |        "2992119  다만 취업 준비를 지원한다는 제도 성격을 고려해 유흥.도박.성인 용품 등 용도나 고...   \n",
119 |        "\n",
120 |        "                                                        id  \n",
121 |        "0                                     8 190 0 42 45 1 1 1   \n",
122 |        "1        318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...  \n",
123 |        "2        3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...  \n",
124 |        "3                                    32 20 79 162 0 6 130   \n",
125 |        "4        233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...  \n",
126 |        "...                                                    ...  \n",
127 |        "2992115  96 0 57 275 0 5 373 0 560 700 350 109 17 0 26 ...  \n",
128 |        "2992116                                            235 47   \n",
129 |        "2992117  351 23 0 449 108 0 26 62 7 0 219 17 194 0 147 ...  \n",
130 |        "2992118                  13 4 0 94 23 0 194 71 216 3 25 1   \n",
131 |        "2992119  15 46 0 331 207 0 245 122 55 0 10 82 27 15 4 0...  \n",
132 |        "\n",
133 |        "[2992120 rows x 2 columns]"
134 |       ]
135 |      },
136 |      "execution_count": 2,
137 |      "metadata": {},
138 |      "output_type": "execute_result"
139 |     }
140 |    ],
141 |    "source": [
142 |     "corpus_df = None\n",
143 |     "\n",
144 |     "with open('corpus_df.bin', 'rb') as f:\n",
145 |     "    corpus_df = pickle.load(f)\n",
146 |     "    \n",
147 |     "corpus_df"
148 |    ]
149 |   },
150 |   {
151 |    "cell_type": "code",
152 |    "execution_count": 4,
153 |    "metadata": {},
154 |    "outputs": [],
155 |    "source": [
156 |     "targets = corpus_df['id']\n",
157 |     "target_lengths = list()\n",
158 |     "\n",
159 |     "for target in targets:\n",
160 |     "    tokens = target.split()\n",
161 |     "    target_lengths.append(len(tokens))"
162 |    ]
163 |   },
164 |   {
165 |    "cell_type": "code",
166 |    "execution_count": 5,
167 |    "metadata": {},
168 |    "outputs": [
169 |     {
170 |      "data": {
171 |       "image/png": "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\n",
172 |       "text/plain": [
173 |        "<Figure size 432x288 with 1 Axes>"
174 |       ]
175 |      },
176 |      "metadata": {
177 |       "needs_background": "light"
178 |      },
179 |      "output_type": "display_data"
180 |     }
181 |    ],
182 |    "source": [
183 |     "sns.boxplot(target_lengths)\n",
184 |     "plt.title('Language Model Corpus length Analysis', fontsize='x-large')\n",
185 |     "plt.xlabel('Sequence Length', fontsize='large')\n",
186 |     "plt.show()"
187 |    ]
188 |   },
189 |   {
190 |    "cell_type": "code",
191 |    "execution_count": 6,
192 |    "metadata": {},
193 |    "outputs": [],
194 |    "source": [
195 |     "sentences = corpus_df['ko']\n",
196 |     "\n",
197 |     "new_sentences = list()\n",
198 |     "new_targets = list()\n",
199 |     "\n",
200 |     "for (sentence, target) in zip(sentences, targets):\n",
201 |     "    if len(target.split()) < 151:\n",
202 |     "        new_sentences.append(sentence)\n",
203 |     "        new_targets.append(target)"
204 |    ]
205 |   },
206 |   {
207 |    "cell_type": "code",
208 |    "execution_count": 8,
209 |    "metadata": {},
210 |    "outputs": [
211 |     {
212 |      "name": "stdout",
213 |      "output_type": "stream",
214 |      "text": [
215 |       "2991682\n",
216 |       "2991682\n"
217 |      ]
218 |     }
219 |    ],
220 |    "source": [
221 |     "print(len(new_sentences))\n",
222 |     "print(len(new_targets))"
223 |    ]
224 |   },
225 |   {
226 |    "cell_type": "code",
227 |    "execution_count": 11,
228 |    "metadata": {},
229 |    "outputs": [
230 |     {
231 |      "data": {
232 |       "text/html": [
233 |        "<div>\n",
234 |        "<style scoped>\n",
235 |        "    .dataframe tbody tr th:only-of-type {\n",
236 |        "        vertical-align: middle;\n",
237 |        "    }\n",
238 |        "\n",
239 |        "    .dataframe tbody tr th {\n",
240 |        "        vertical-align: top;\n",
241 |        "    }\n",
242 |        "\n",
243 |        "    .dataframe thead th {\n",
244 |        "        text-align: right;\n",
245 |        "    }\n",
246 |        "</style>\n",
247 |        "<table border=\"1\" class=\"dataframe\">\n",
248 |        "  <thead>\n",
249 |        "    <tr style=\"text-align: right;\">\n",
250 |        "      <th></th>\n",
251 |        "      <th>ko</th>\n",
252 |        "      <th>id</th>\n",
253 |        "    </tr>\n",
254 |        "  </thead>\n",
255 |        "  <tbody>\n",
256 |        "    <tr>\n",
257 |        "      <th>0</th>\n",
258 |        "      <td>어디 보자...</td>\n",
259 |        "      <td>8 190 0 42 45 1 1 1</td>\n",
260 |        "    </tr>\n",
261 |        "    <tr>\n",
262 |        "      <th>1</th>\n",
263 |        "      <td>칠대 왕국 종족 중에 오크는 처음 듣는다</td>\n",
264 |        "      <td>318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...</td>\n",
265 |        "    </tr>\n",
266 |        "    <tr>\n",
267 |        "      <th>2</th>\n",
268 |        "      <td>이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.</td>\n",
269 |        "      <td>3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...</td>\n",
270 |        "    </tr>\n",
271 |        "    <tr>\n",
272 |        "      <th>3</th>\n",
273 |        "      <td>라니스터 가의</td>\n",
274 |        "      <td>32 20 79 162 0 6 130</td>\n",
275 |        "    </tr>\n",
276 |        "    <tr>\n",
277 |        "      <th>4</th>\n",
278 |        "      <td>별 희한한 생각이 다 떠오르곤 하죠</td>\n",
279 |        "      <td>233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...</td>\n",
280 |        "    </tr>\n",
281 |        "  </tbody>\n",
282 |        "</table>\n",
283 |        "</div>"
284 |       ],
285 |       "text/plain": [
286 |        "                                           ko  \\\n",
287 |        "0                                    어디 보자...   \n",
288 |        "1                      칠대 왕국 종족 중에 오크는 처음 듣는다   \n",
289 |        "2  이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.   \n",
290 |        "3                                     라니스터 가의   \n",
291 |        "4                         별 희한한 생각이 다 떠오르곤 하죠   \n",
292 |        "\n",
293 |        "                                                  id  \n",
294 |        "0                               8 190 0 42 45 1 1 1   \n",
295 |        "1  318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...  \n",
296 |        "2  3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...  \n",
297 |        "3                              32 20 79 162 0 6 130   \n",
298 |        "4  233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...  "
299 |       ]
300 |      },
301 |      "execution_count": 11,
302 |      "metadata": {},
303 |      "output_type": "execute_result"
304 |     }
305 |    ],
306 |    "source": [
307 |     "corpus_dict = {'ko' : new_sentences,\n",
308 |     "               'id' : new_targets}\n",
309 |     "corpus_df = pd.DataFrame(corpus_dict)\n",
310 |     "corpus_df.head()"
311 |    ]
312 |   },
313 |   {
314 |    "cell_type": "code",
315 |    "execution_count": 12,
316 |    "metadata": {},
317 |    "outputs": [],
318 |    "source": [
319 |     "with open('corpus_df.bin', 'wb') as f:\n",
320 |     "    pickle.dump(corpus_df, f)"
321 |    ]
322 |   },
323 |   {
324 |    "cell_type": "code",
325 |    "execution_count": null,
326 |    "metadata": {},
327 |    "outputs": [],
328 |    "source": []
329 |   }
330 |  ],
331 |  "metadata": {
332 |   "kernelspec": {
333 |    "display_name": "Python 3",
334 |    "language": "python",
335 |    "name": "python3"
336 |   },
337 |   "language_info": {
338 |    "codemirror_mode": {
339 |     "name": "ipython",
340 |     "version": 3
341 |    },
342 |    "file_extension": ".py",
343 |    "mimetype": "text/x-python",
344 |    "name": "python",
345 |    "nbconvert_exporter": "python",
346 |    "pygments_lexer": "ipython3",
347 |    "version": "3.7.6"
348 |   }
349 |  },
350 |  "nbformat": 4,
351 |  "nbformat_minor": 4
352 | }
353 | 


--------------------------------------------------------------------------------
/data/Data-Analysis.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "metadata": {},
  7 |    "outputs": [],
  8 |    "source": [
  9 |     "import pickle\n",
 10 |     "import pandas as pd\n",
 11 |     "import matplotlib.pyplot as plt\n",
 12 |     "import seaborn as sns"
 13 |    ]
 14 |   },
 15 |   {
 16 |    "cell_type": "code",
 17 |    "execution_count": 2,
 18 |    "metadata": {},
 19 |    "outputs": [
 20 |     {
 21 |      "data": {
 22 |       "text/html": [
 23 |        "<div>\n",
 24 |        "<style scoped>\n",
 25 |        "    .dataframe tbody tr th:only-of-type {\n",
 26 |        "        vertical-align: middle;\n",
 27 |        "    }\n",
 28 |        "\n",
 29 |        "    .dataframe tbody tr th {\n",
 30 |        "        vertical-align: top;\n",
 31 |        "    }\n",
 32 |        "\n",
 33 |        "    .dataframe thead th {\n",
 34 |        "        text-align: right;\n",
 35 |        "    }\n",
 36 |        "</style>\n",
 37 |        "<table border=\"1\" class=\"dataframe\">\n",
 38 |        "  <thead>\n",
 39 |        "    <tr style=\"text-align: right;\">\n",
 40 |        "      <th></th>\n",
 41 |        "      <th>ko</th>\n",
 42 |        "      <th>id</th>\n",
 43 |        "    </tr>\n",
 44 |        "  </thead>\n",
 45 |        "  <tbody>\n",
 46 |        "    <tr>\n",
 47 |        "      <th>0</th>\n",
 48 |        "      <td>어디 보자...</td>\n",
 49 |        "      <td>8 190 0 42 45 1 1 1</td>\n",
 50 |        "    </tr>\n",
 51 |        "    <tr>\n",
 52 |        "      <th>1</th>\n",
 53 |        "      <td>칠대 왕국 종족 중에 오크는 처음 듣는다</td>\n",
 54 |        "      <td>318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...</td>\n",
 55 |        "    </tr>\n",
 56 |        "    <tr>\n",
 57 |        "      <th>2</th>\n",
 58 |        "      <td>이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.</td>\n",
 59 |        "      <td>3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...</td>\n",
 60 |        "    </tr>\n",
 61 |        "    <tr>\n",
 62 |        "      <th>3</th>\n",
 63 |        "      <td>라니스터 가의</td>\n",
 64 |        "      <td>32 20 79 162 0 6 130</td>\n",
 65 |        "    </tr>\n",
 66 |        "    <tr>\n",
 67 |        "      <th>4</th>\n",
 68 |        "      <td>별 희한한 생각이 다 떠오르곤 하죠</td>\n",
 69 |        "      <td>233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...</td>\n",
 70 |        "    </tr>\n",
 71 |        "    <tr>\n",
 72 |        "      <th>...</th>\n",
 73 |        "      <td>...</td>\n",
 74 |        "      <td>...</td>\n",
 75 |        "    </tr>\n",
 76 |        "    <tr>\n",
 77 |        "      <th>2992115</th>\n",
 78 |        "      <td>네 오늘 아침 범죄현장에 있었죠</td>\n",
 79 |        "      <td>96 0 57 275 0 5 373 0 560 700 350 109 17 0 26 ...</td>\n",
 80 |        "    </tr>\n",
 81 |        "    <tr>\n",
 82 |        "      <th>2992116</th>\n",
 83 |        "      <td>머시</td>\n",
 84 |        "      <td>235 47</td>\n",
 85 |        "    </tr>\n",
 86 |        "    <tr>\n",
 87 |        "      <th>2992117</th>\n",
 88 |        "      <td>눈은 풀려 있었고 입에선 연신 침이 흘러 나왔다.</td>\n",
 89 |        "      <td>351 23 0 449 108 0 26 62 7 0 219 17 194 0 147 ...</td>\n",
 90 |        "    </tr>\n",
 91 |        "    <tr>\n",
 92 |        "      <th>2992118</th>\n",
 93 |        "      <td>나는 좋은 선생님이야.</td>\n",
 94 |        "      <td>13 4 0 94 23 0 194 71 216 3 25 1</td>\n",
 95 |        "    </tr>\n",
 96 |        "    <tr>\n",
 97 |        "      <th>2992119</th>\n",
 98 |        "      <td>다만 취업 준비를 지원한다는 제도 성격을 고려해 유흥.도박.성인 용품 등 용도나 고...</td>\n",
 99 |        "      <td>15 46 0 331 207 0 245 122 55 0 10 82 27 15 4 0...</td>\n",
100 |        "    </tr>\n",
101 |        "  </tbody>\n",
102 |        "</table>\n",
103 |        "<p>2992120 rows × 2 columns</p>\n",
104 |        "</div>"
105 |       ],
106 |       "text/plain": [
107 |        "                                                        ko  \\\n",
108 |        "0                                                 어디 보자...   \n",
109 |        "1                                   칠대 왕국 종족 중에 오크는 처음 듣는다   \n",
110 |        "2               이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.   \n",
111 |        "3                                                  라니스터 가의   \n",
112 |        "4                                      별 희한한 생각이 다 떠오르곤 하죠   \n",
113 |        "...                                                    ...   \n",
114 |        "2992115                                  네 오늘 아침 범죄현장에 있었죠   \n",
115 |        "2992116                                                 머시   \n",
116 |        "2992117                        눈은 풀려 있었고 입에선 연신 침이 흘러 나왔다.   \n",
117 |        "2992118                                       나는 좋은 선생님이야.   \n",
118 |        "2992119  다만 취업 준비를 지원한다는 제도 성격을 고려해 유흥.도박.성인 용품 등 용도나 고...   \n",
119 |        "\n",
120 |        "                                                        id  \n",
121 |        "0                                     8 190 0 42 45 1 1 1   \n",
122 |        "1        318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...  \n",
123 |        "2        3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...  \n",
124 |        "3                                    32 20 79 162 0 6 130   \n",
125 |        "4        233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...  \n",
126 |        "...                                                    ...  \n",
127 |        "2992115  96 0 57 275 0 5 373 0 560 700 350 109 17 0 26 ...  \n",
128 |        "2992116                                            235 47   \n",
129 |        "2992117  351 23 0 449 108 0 26 62 7 0 219 17 194 0 147 ...  \n",
130 |        "2992118                  13 4 0 94 23 0 194 71 216 3 25 1   \n",
131 |        "2992119  15 46 0 331 207 0 245 122 55 0 10 82 27 15 4 0...  \n",
132 |        "\n",
133 |        "[2992120 rows x 2 columns]"
134 |       ]
135 |      },
136 |      "execution_count": 2,
137 |      "metadata": {},
138 |      "output_type": "execute_result"
139 |     }
140 |    ],
141 |    "source": [
142 |     "corpus_df = None\n",
143 |     "\n",
144 |     "with open('corpus_df.bin', 'rb') as f:\n",
145 |     "    corpus_df = pickle.load(f)\n",
146 |     "    \n",
147 |     "corpus_df"
148 |    ]
149 |   },
150 |   {
151 |    "cell_type": "code",
152 |    "execution_count": 4,
153 |    "metadata": {},
154 |    "outputs": [],
155 |    "source": [
156 |     "targets = corpus_df['id']\n",
157 |     "target_lengths = list()\n",
158 |     "\n",
159 |     "for target in targets:\n",
160 |     "    tokens = target.split()\n",
161 |     "    target_lengths.append(len(tokens))"
162 |    ]
163 |   },
164 |   {
165 |    "cell_type": "code",
166 |    "execution_count": 5,
167 |    "metadata": {},
168 |    "outputs": [
169 |     {
170 |      "data": {
171 |       "image/png": "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\n",
172 |       "text/plain": [
173 |        "<Figure size 432x288 with 1 Axes>"
174 |       ]
175 |      },
176 |      "metadata": {
177 |       "needs_background": "light"
178 |      },
179 |      "output_type": "display_data"
180 |     }
181 |    ],
182 |    "source": [
183 |     "sns.boxplot(target_lengths)\n",
184 |     "plt.title('Language Model Corpus length Analysis', fontsize='x-large')\n",
185 |     "plt.xlabel('Sequence Length', fontsize='large')\n",
186 |     "plt.show()"
187 |    ]
188 |   },
189 |   {
190 |    "cell_type": "code",
191 |    "execution_count": 6,
192 |    "metadata": {},
193 |    "outputs": [],
194 |    "source": [
195 |     "sentences = corpus_df['ko']\n",
196 |     "\n",
197 |     "new_sentences = list()\n",
198 |     "new_targets = list()\n",
199 |     "\n",
200 |     "for (sentence, target) in zip(sentences, targets):\n",
201 |     "    if len(target.split()) < 151:\n",
202 |     "        new_sentences.append(sentence)\n",
203 |     "        new_targets.append(target)"
204 |    ]
205 |   },
206 |   {
207 |    "cell_type": "code",
208 |    "execution_count": 8,
209 |    "metadata": {},
210 |    "outputs": [
211 |     {
212 |      "name": "stdout",
213 |      "output_type": "stream",
214 |      "text": [
215 |       "2991682\n",
216 |       "2991682\n"
217 |      ]
218 |     }
219 |    ],
220 |    "source": [
221 |     "print(len(new_sentences))\n",
222 |     "print(len(new_targets))"
223 |    ]
224 |   },
225 |   {
226 |    "cell_type": "code",
227 |    "execution_count": 11,
228 |    "metadata": {},
229 |    "outputs": [
230 |     {
231 |      "data": {
232 |       "text/html": [
233 |        "<div>\n",
234 |        "<style scoped>\n",
235 |        "    .dataframe tbody tr th:only-of-type {\n",
236 |        "        vertical-align: middle;\n",
237 |        "    }\n",
238 |        "\n",
239 |        "    .dataframe tbody tr th {\n",
240 |        "        vertical-align: top;\n",
241 |        "    }\n",
242 |        "\n",
243 |        "    .dataframe thead th {\n",
244 |        "        text-align: right;\n",
245 |        "    }\n",
246 |        "</style>\n",
247 |        "<table border=\"1\" class=\"dataframe\">\n",
248 |        "  <thead>\n",
249 |        "    <tr style=\"text-align: right;\">\n",
250 |        "      <th></th>\n",
251 |        "      <th>ko</th>\n",
252 |        "      <th>id</th>\n",
253 |        "    </tr>\n",
254 |        "  </thead>\n",
255 |        "  <tbody>\n",
256 |        "    <tr>\n",
257 |        "      <th>0</th>\n",
258 |        "      <td>어디 보자...</td>\n",
259 |        "      <td>8 190 0 42 45 1 1 1</td>\n",
260 |        "    </tr>\n",
261 |        "    <tr>\n",
262 |        "      <th>1</th>\n",
263 |        "      <td>칠대 왕국 종족 중에 오크는 처음 듣는다</td>\n",
264 |        "      <td>318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...</td>\n",
265 |        "    </tr>\n",
266 |        "    <tr>\n",
267 |        "      <th>2</th>\n",
268 |        "      <td>이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.</td>\n",
269 |        "      <td>3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...</td>\n",
270 |        "    </tr>\n",
271 |        "    <tr>\n",
272 |        "      <th>3</th>\n",
273 |        "      <td>라니스터 가의</td>\n",
274 |        "      <td>32 20 79 162 0 6 130</td>\n",
275 |        "    </tr>\n",
276 |        "    <tr>\n",
277 |        "      <th>4</th>\n",
278 |        "      <td>별 희한한 생각이 다 떠오르곤 하죠</td>\n",
279 |        "      <td>233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...</td>\n",
280 |        "    </tr>\n",
281 |        "  </tbody>\n",
282 |        "</table>\n",
283 |        "</div>"
284 |       ],
285 |       "text/plain": [
286 |        "                                           ko  \\\n",
287 |        "0                                    어디 보자...   \n",
288 |        "1                      칠대 왕국 종족 중에 오크는 처음 듣는다   \n",
289 |        "2  이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.   \n",
290 |        "3                                     라니스터 가의   \n",
291 |        "4                         별 희한한 생각이 다 떠오르곤 하죠   \n",
292 |        "\n",
293 |        "                                                  id  \n",
294 |        "0                               8 190 0 42 45 1 1 1   \n",
295 |        "1  318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...  \n",
296 |        "2  3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...  \n",
297 |        "3                              32 20 79 162 0 6 130   \n",
298 |        "4  233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...  "
299 |       ]
300 |      },
301 |      "execution_count": 11,
302 |      "metadata": {},
303 |      "output_type": "execute_result"
304 |     }
305 |    ],
306 |    "source": [
307 |     "corpus_dict = {'ko' : new_sentences,\n",
308 |     "               'id' : new_targets}\n",
309 |     "corpus_df = pd.DataFrame(corpus_dict)\n",
310 |     "corpus_df.head()"
311 |    ]
312 |   },
313 |   {
314 |    "cell_type": "code",
315 |    "execution_count": 12,
316 |    "metadata": {},
317 |    "outputs": [],
318 |    "source": [
319 |     "with open('corpus_df.bin', 'wb') as f:\n",
320 |     "    pickle.dump(corpus_df, f)"
321 |    ]
322 |   },
323 |   {
324 |    "cell_type": "code",
325 |    "execution_count": null,
326 |    "metadata": {},
327 |    "outputs": [],
328 |    "source": []
329 |   }
330 |  ],
331 |  "metadata": {
332 |   "kernelspec": {
333 |    "display_name": "Python 3",
334 |    "language": "python",
335 |    "name": "python3"
336 |   },
337 |   "language_info": {
338 |    "codemirror_mode": {
339 |     "name": "ipython",
340 |     "version": 3
341 |    },
342 |    "file_extension": ".py",
343 |    "mimetype": "text/x-python",
344 |    "name": "python",
345 |    "nbconvert_exporter": "python",
346 |    "pygments_lexer": "ipython3",
347 |    "version": "3.7.6"
348 |   }
349 |  },
350 |  "nbformat": 4,
351 |  "nbformat_minor": 4
352 | }
353 | 


--------------------------------------------------------------------------------
/data/train_labels.csv:
--------------------------------------------------------------------------------
   1 | id,char,freq
   2 | 0, ,5774462
   3 | 1,.,640924
   4 | 2,그,556373
   5 | 3,이,509291
   6 | 4,는,374559
   7 | 5,아,370444
   8 | 6,가,369698
   9 | 7,고,356378
  10 | 8,어,333842
  11 | 9,거,306987
  12 | 10,지,276453
  13 | 11,데,249269
  14 | 12,?,235024
  15 | 13,나,229646
  16 | 14,하,226136
  17 | 15,다,221216
  18 | 16,서,211193
  19 | 17,에,204330
  20 | 18,도,190561
  21 | 19,게,177140
  22 | 20,니,173284
  23 | 21,",",152938
  24 | 22,기,149467
  25 | 23,은,144674
  26 | 24,면,142025
  27 | 25,야,137553
  28 | 26,있,133155
  29 | 27,한,121564
  30 | 28,을,121048
  31 | 29,까,119483
  32 | 30,해,115148
  33 | 31,리,111855
  34 | 32,라,111479
  35 | 33,래,105784
  36 | 34,사,100533
  37 | 35,근,99781
  38 | 36,들,99447
  39 | 37,안,97043
  40 | 38,로,91847
  41 | 39,일,88319
  42 | 40,뭐,87328
  43 | 41,내,85968
  44 | 42,보,82911
  45 | 43,제,80874
  46 | 44,같,79626
  47 | 45,자,76298
  48 | 46,만,76093
  49 | 47,시,72836
  50 | 48,런,70919
  51 | 49,너,69192
  52 | 50,대,68756
  53 | 51,때,67179
  54 | 52,되,66237
  55 | 53,으,66106
  56 | 54,진,62831
  57 | 55,를,61802
  58 | 56,잖,61455
  59 | 57,오,60782
  60 | 58,러,60629
  61 | 59,인,60234
  62 | 60,막,59994
  63 | 61,무,58705
  64 | 62,었,58385
  65 | 63,구,57294
  66 | 64,했,57209
  67 | 65,수,56787
  68 | 66,간,55275
  69 | 67,애,54476
  70 | 68,우,53539
  71 | 69,요,53234
  72 | 70,마,53125
  73 | 71,생,52815
  74 | 72,렇,50798
  75 | 73,냥,49989
  76 | 74,짜,49581
  77 | 75,주,48969
  78 | 76,없,48392
  79 | 77,말,47929
  80 | 78,학,46285
  81 | 79,스,46225
  82 | 80,더,44487
  83 | 81,많,43607
  84 | 82,원,41379
  85 | 83,음,41348
  86 | 84,정,39775
  87 | 85,겠,39691
  88 | 86,여,39203
  89 | 87,먹,39194
  90 | 88,금,38720
  91 | 89,든,38476
  92 | 90,부,38398
  93 | 91,할,38262
  94 | 92,전,36575
  95 | 93,번,36375
  96 | 94,좋,36363
  97 | 95,랑,36081
  98 | 96,네,35514
  99 | 97,람,33799
 100 | 98,약,33412
 101 | 99,건,33371
 102 | 100,각,32167
 103 | 101,좀,31738
 104 | 102,알,30893
 105 | 103,잘,30132
 106 | 104,걸,29634
 107 | 105,모,29629
 108 | 106,것,28482
 109 | 107,상,28247
 110 | 108,려,28218
 111 | 109,장,27856
 112 | 110,히,27705
 113 | 111,않,27305
 114 | 112,맞,27202
 115 | 113,던,27082
 116 | 114,르,26286
 117 | 115,교,26116
 118 | 116,바,25994
 119 | 117,냐,25742
 120 | 118,드,25702
 121 | 119,십,25654
 122 | 120,날,25556
 123 | 121,치,25287
 124 | 122,비,25278
 125 | 123,단,25129
 126 | 124,동,25047
 127 | 125,또,24720
 128 | 126,못,24528
 129 | 127,저,24074
 130 | 128,얘,23990
 131 | 129,중,23851
 132 | 130,의,23607
 133 | 131,난,23318
 134 | 132,엄,23057
 135 | 133,봤,22930
 136 | 134,걔,22732
 137 | 135,화,22593
 138 | 136,응,22254
 139 | 137,싶,21756
 140 | 138,갔,21628
 141 | 139,았,21052
 142 | 140,집,20850
 143 | 141,왜,20801
 144 | 142,계,20757
 145 | 143,공,20620
 146 | 144,긴,20547
 147 | 145,신,20371
 148 | 146,적,20244
 149 | 147,연,20225
 150 | 148,직,20061
 151 | 149,실,19467
 152 | 150,영,19454
 153 | 151,미,19444
 154 | 152,봐,18931
 155 | 153,분,18893
 156 | 154,테,18829
 157 | 155,년,18669
 158 | 156,트,18654
 159 | 157,문,18230
 160 | 158,와,18114
 161 | 159,돼,18028
 162 | 160,물,17889
 163 | 161,예,17864
 164 | 162,터,17722
 165 | 163,세,17719
 166 | 164,럼,17521
 167 | 165,청,17479
 168 | 166,차,17455
 169 | 167,친,17355
 170 | 168,개,17355
 171 | 169,삼,17242
 172 | 170,국,17224
 173 | 171,두,17129
 174 | 172,소,17125
 175 | 173,살,16893
 176 | 174,재,16635
 177 | 175,운,15949
 178 | 176,쫌,15780
 179 | 177,유,15516
 180 | 178,속,15255
 181 | 179,명,15190
 182 | 180,랬,15155
 183 | 181,본,15148
 184 | 182,갈,15084
 185 | 183,방,15069
 186 | 184,돈,14998
 187 | 185,타,14919
 188 | 186,처,14908
 189 | 187,빠,14851
 190 | 188,과,14843
 191 | 189,식,14743
 192 | 190,디,14633
 193 | 191,배,14403
 194 | 192,피,14093
 195 | 193,뭔,14093
 196 | 194,선,13929
 197 | 195,남,13909
 198 | 196,경,13868
 199 | 197,달,13787
 200 | 198,언,13579
 201 | 199,받,13519
 202 | 200,심,13409
 203 | 201,월,13367
 204 | 202,버,13339
 205 | 203,왔,13265
 206 | 204,느,13223
 207 | 205,점,12960
 208 | 206,올,12923
 209 | 207,업,12861
 210 | 208,른,12801
 211 | 209,성,12717
 212 | 210,회,12591
 213 | 211,조,12570
 214 | 212,워,12424
 215 | 213,따,12410
 216 | 214,행,12350
 217 | 215,반,12158
 218 | 216,님,11998
 219 | 217,딱,11908
 220 | 218,관,11828
 221 | 219,입,11275
 222 | 220,카,11235
 223 | 221,당,11068
 224 | 222,였,10977
 225 | 223,케,10576
 226 | 224,쪽,10430
 227 | 225,천,10384
 228 | 226,작,10381
 229 | 227,누,10336
 230 | 228,열,10252
 231 | 229,얼,10250
 232 | 230,울,10246
 233 | 231,찮,10231
 234 | 232,된,10191
 235 | 233,별,10159
 236 | 234,떻,10108
 237 | 235,머,9876
 238 | 236,쓰,9853
 239 | 237,위,9841
 240 | 238,크,9838
 241 | 239,노,9799
 242 | 240,괜,9735
 243 | 241,강,9698
 244 | 242,편,9668
 245 | 243,몰,9623
 246 | 244,맛,9382
 247 | 245,준,9342
 248 | 246,줄,9294
 249 | 247,파,9282
 250 | 248,백,9252
 251 | 249,매,9181
 252 | 250,산,9160
 253 | 251,술,9142
 254 | 252,힘,9056
 255 | 253,프,9019
 256 | 254,즘,8997
 257 | 255,임,8969
 258 | 256,체,8888
 259 | 257,형,8790
 260 | 258,몇,8742
 261 | 259,맨,8712
 262 | 260,새,8623
 263 | 261,용,8571
 264 | 262,키,8547
 265 | 263,통,8410
 266 | 264,양,8363
 267 | 265,끝,8361
 268 | 266,싸,8328
 269 | 267,볼,8188
 270 | 268,혼,8179
 271 | 269,온,8132
 272 | 270,등,8123
 273 | 271,길,8067
 274 | 272,될,8033
 275 | 273,밌,7998
 276 | 274,육,7924
 277 | 275,늘,7909
 278 | 276,슨,7835
 279 | 277,됐,7738
 280 | 278,놀,7707
 281 | 279,외,7608
 282 | 280,팔,7601
 283 | 281,져,7551
 284 | 282,레,7485
 285 | 283,억,7461
 286 | 284,발,7450
 287 | 285,결,7412
 288 | 286,초,7290
 289 | 287,감,7180
 290 | 288,군,7174
 291 | 289,호,7173
 292 | 290,름,7146
 293 | 291,솔,7079
 294 | 292,닌,7051
 295 | 293,밖,7013
 296 | 294,불,7007
 297 | 295,밥,6784
 298 | 296,포,6676
 299 | 297,싫,6631
 300 | 298,완,6582
 301 | 299,갖,6511
 302 | 300,겨,6468
 303 | 301,질,6453
 304 | 302,토,6448
 305 | 303,험,6417
 306 | 304,색,6371
 307 | 305,떤,6352
 308 | 306,역,6340
 309 | 307,티,6319
 310 | 308,갑,6316
 311 | 309,목,6262
 312 | 310,린,6256
 313 | 311,추,6204
 314 | 312,격,6174
 315 | 313,후,6119
 316 | 314,확,6095
 317 | 315,루,6079
 318 | 316,민,6024
 319 | 317,끼,6023
 320 | 318,칠,6019
 321 | 319,돌,5997
 322 | 320,찍,5956
 323 | 321,쪼,5946
 324 | 322,깐,5788
 325 | 323,필,5786
 326 | 324,빨,5693
 327 | 325,났,5657
 328 | 326,락,5561
 329 | 327,박,5537
 330 | 328,끔,5483
 331 | 329,낌,5403
 332 | 330,럴,5397
 333 | 331,취,5344
 334 | 332,복,5315
 335 | 333,둘,5264
 336 | 334,페,5217
 337 | 335,렸,5198
 338 | 336,써,5197
 339 | 337,줘,5173
 340 | 338,급,5067
 341 | 339,력,5065
 342 | 340,잡,5030
 343 | 341,씩,5006
 344 | 342,찾,4990
 345 | 343,놓,4987
 346 | 344,최,4894
 347 | 345,코,4891
 348 | 346,넘,4870
 349 | 347,졌,4803
 350 | 348,섯,4799
 351 | 349,브,4793
 352 | 350,현,4764
 353 | 351,눈,4760
 354 | 352,항,4751
 355 | 353,귀,4708
 356 | 354,설,4688
 357 | 355,벌,4666
 358 | 356,담,4647
 359 | 357,앞,4640
 360 | 358,책,4630
 361 | 359,절,4629
 362 | 360,플,4523
 363 | 361,폰,4513
 364 | 362,태,4496
 365 | 363,종,4487
 366 | 364,옛,4450
 367 | 365,증,4413
 368 | 366,튼,4411
 369 | 367,글,4408
 370 | 368,습,4383
 371 | 369,병,4377
 372 | 370,론,4373
 373 | 371,출,4364
 374 | 372,능,4354
 375 | 373,침,4345
 376 | 374,순,4339
 377 | 375,줬,4308
 378 | 376,평,4303
 379 | 377,메,4287
 380 | 378,똑,4281
 381 | 379,커,4261
 382 | 380,엔,4248
 383 | 381,꾸,4230
 384 | 382,란,4194
 385 | 383,듣,4083
 386 | 384,씨,4009
 387 | 385,큰,4002
 388 | 386,표,3995
 389 | 387,잠,3942
 390 | 388,먼,3942
 391 | 389,쁘,3840
 392 | 390,활,3820
 393 | 391,합,3787
 394 | 392,접,3732
 395 | 393,럽,3722
 396 | 394,옷,3705
 397 | 395,쳐,3690
 398 | 396,손,3689
 399 | 397,붙,3645
 400 | 398,망,3640
 401 | 399,죽,3609
 402 | 400,투,3606
 403 | 401,족,3603
 404 | 402,셨,3589
 405 | 403,참,3572
 406 | 404,떨,3567
 407 | 405,웃,3533
 408 | 406,졸,3516
 409 | 407,쉬,3492
 410 | 408,뭘,3447
 411 | 409,변,3406
 412 | 410,릴,3374
 413 | 411,웠,3293
 414 | 412,홍,3267
 415 | 413,즈,3265
 416 | 414,랐,3245
 417 | 415,독,3243
 418 | 416,충,3239
 419 | 417,짝,3217
 420 | 418,떡,3197
 421 | 419,뒤,3195
 422 | 420,휴,3161
 423 | 421,셔,3142
 424 | 422,넣,3135
 425 | 423,쨌,3075
 426 | 424,악,3073
 427 | 425,패,3049
 428 | 426,빼,3041
 429 | 427,슬,2983
 430 | 428,특,2975
 431 | 429,꺼,2970
 432 | 430,숙,2951
 433 | 431,쯤,2934
 434 | 432,텐,2905
 435 | 433,창,2901
 436 | 434,겼,2888
 437 | 435,굴,2869
 438 | 436,판,2863
 439 | 437,죠,2851
 440 | 438,답,2820
 441 | 439,희,2816
 442 | 440,허,2815
 443 | 441,옆,2798
 444 | 442,료,2791
 445 | 443,닐,2790
 446 | 444,택,2769
 447 | 445,림,2760
 448 | 446,읽,2742
 449 | 447,핸,2733
 450 | 448,축,2730
 451 | 449,풀,2716
 452 | 450,틀,2712
 453 | 451,몸,2694
 454 | 452,골,2690
 455 | 453,황,2635
 456 | 454,켜,2635
 457 | 455,익,2625
 458 | 456,베,2613
 459 | 457,북,2600
 460 | 458,법,2578
 461 | 459,늦,2578
 462 | 460,함,2568
 463 | 461,랜,2555
 464 | 462,꼬,2555
 465 | 463,향,2547
 466 | 464,석,2541
 467 | 465,환,2533
 468 | 466,슷,2529
 469 | 467,품,2518
 470 | 468,혀,2513
 471 | 469,블,2512
 472 | 470,쓸,2503
 473 | 471,채,2472
 474 | 472,며,2470
 475 | 473,욕,2463
 476 | 474,권,2450
 477 | 475,검,2445
 478 | 476,굳,2428
 479 | 477,록,2425
 480 | 478,톡,2408
 481 | 479,김,2408
 482 | 480,넌,2383
 483 | 481,깨,2375
 484 | 482,션,2374
 485 | 483,캐,2369
 486 | 484,송,2339
 487 | 485,녀,2336
 488 | 486,탈,2327
 489 | 487,광,2321
 490 | 488,혹,2313
 491 | 489,퍼,2300
 492 | 490,뽑,2278
 493 | 491,철,2265
 494 | 492,째,2249
 495 | 493,움,2232
 496 | 494,밤,2231
 497 | 495,꼭,2226
 498 | 496,샀,2224
 499 | 497,끊,2212
 500 | 498,땐,2203
 501 | 499,깔,2179
 502 | 500,멀,2145
 503 | 501,높,2141
 504 | 502,께,2140
 505 | 503,큼,2121
 506 | 504,녁,2104
 507 | 505,곳,2082
 508 | 506,잔,2070
 509 | 507,쉽,2070
 510 | 508,짐,2067
 511 | 509,암,2063
 512 | 510,극,2061
 513 | 511,련,2056
 514 | 512,떠,2056
 515 | 513,벽,2049
 516 | 514,헤,2047
 517 | 515,C,2040
 518 | 516,끄,2024
 519 | 517,곱,2015
 520 | 518,승,2011
 521 | 519,봉,2009
 522 | 520,착,2006
 523 | 521,촌,1990
 524 | 522,껴,1986
 525 | 523,딩,1983
 526 | 524,류,1978
 527 | 525,뜨,1970
 528 | 526,넷,1941
 529 | 527,놨,1922
 530 | 528,궁,1894
 531 | 529,논,1882
 532 | 530,곤,1875
 533 | 531,클,1869
 534 | 532,싼,1859
 535 | 533,앉,1854
 536 | 534,컴,1849
 537 | 535,맥,1841
 538 | 536,팀,1830
 539 | 537,썼,1818
 540 | 538,낫,1801
 541 | 539,튜,1788
 542 | 540,걱,1786
 543 | 541,쁜,1770
 544 | 542,킨,1760
 545 | 543,빌,1752
 546 | 544,쿠,1748
 547 | 545,찌,1738
 548 | 546,쌤,1719
 549 | 547,T,1715
 550 | 548,밀,1711
 551 | 549,빵,1702
 552 | 550,냈,1702
 553 | 551,센,1691
 554 | 552,딴,1688
 555 | 553,쩌,1678
 556 | 554,딸,1678
 557 | 555,걍,1596
 558 | 556,획,1588
 559 | 557,씬,1582
 560 | 558,챙,1541
 561 | 559,첫,1536
 562 | 560,범,1530
 563 | 561,핑,1519
 564 | 562,굉,1519
 565 | 563,쩔,1514
 566 | 564,팅,1507
 567 | 565,긍,1486
 568 | 566,탄,1471
 569 | 567,덟,1470
 570 | 568,퇴,1469
 571 | 569,뛰,1469
 572 | 570,층,1467
 573 | 571,춰,1454
 574 | 572,훨,1447
 575 | 573,찬,1439
 576 | 574,듯,1424
 577 | 575,S,1396
 578 | 576,왕,1392
 579 | 577,텔,1385
 580 | 578,뉴,1382
 581 | 579,렌,1377
 582 | 580,탕,1374
 583 | 581,짓,1371
 584 | 582,밑,1365
 585 | 583,헬,1358
 586 | 584,존,1339
 587 | 585,립,1323
 588 | 586,녔,1318
 589 | 587,꼈,1305
 590 | 588,빡,1304
 591 | 589,낮,1287
 592 | 590,견,1282
 593 | 591,링,1281
 594 | 592,볶,1271
 595 | 593,낙,1271
 596 | 594,릭,1267
 597 | 595,젠,1263
 598 | 596,퓨,1262
 599 | 597,츠,1256
 600 | 598,맘,1252
 601 | 599,놔,1247
 602 | 600,렵,1241
 603 | 601,땜,1232
 604 | 602,쇼,1224
 605 | 603,값,1215
 606 | 604,닭,1203
 607 | 605,깝,1200
 608 | 606,픈,1194
 609 | 607,탁,1183
 610 | 608,쓴,1179
 611 | 609,농,1172
 612 | 610,량,1166
 613 | 611,염,1156
 614 | 612,홉,1144
 615 | 613,척,1130
 616 | 614,겁,1129
 617 | 615,콘,1127
 618 | 616,섭,1125
 619 | 617,냄,1125
 620 | 618,P,1125
 621 | 619,효,1124
 622 | 620,규,1124
 623 | 621,꿈,1121
 624 | 622,곡,1093
 625 | 623,액,1090
 626 | 624,쎄,1077
 627 | 625,덜,1075
 628 | 626,턴,1065
 629 | 627,킹,1061
 630 | 628,훈,1057
 631 | 629,쳤,1054
 632 | 630,널,1047
 633 | 631,멋,1037
 634 | 632,꿀,1034
 635 | 633,깜,1019
 636 | 634,짧,1014
 637 | 635,롤,1013
 638 | 636,낼,1012
 639 | 637,꽤,1003
 640 | 638,총,984
 641 | 639,램,984
 642 | 640,덕,980
 643 | 641,믄,974
 644 | 642,믿,972
 645 | 643,흥,970
 646 | 644,롱,967
 647 | 645,뜻,962
 648 | 646,짤,958
 649 | 647,쌍,957
 650 | 648,컨,953
 651 | 649,셋,952
 652 | 650,잤,950
 653 | 651,닥,950
 654 | 652,웬,946
 655 | 653,엽,944
 656 | 654,혜,939
 657 | 655,찰,935
 658 | 656,뻐,935
 659 | 657,뿌,934
 660 | 658,빈,934
 661 | 659,꿔,934
 662 | 660,낸,932
 663 | 661,뻔,928
 664 | 662,쌓,926
 665 | 663,즐,919
 666 | 664,튀,914
 667 | 665,겹,909
 668 | 666,득,899
 669 | 667,끌,896
 670 | 668,M,880
 671 | 669,V,877
 672 | 670,녹,876
 673 | 671,푸,870
 674 | 672,쭉,863
 675 | 673,싱,858
 676 | 674,팬,857
 677 | 675,A,854
 678 | 676,!,841
 679 | 677,념,836
 680 | 678,맡,825
 681 | 679,쟁,814
 682 | 680,엑,810
 683 | 681,켓,809
 684 | 682,뀌,808
 685 | 683,털,803
 686 | 684,풍,802
 687 | 685,웨,799
 688 | 686,땡,792
 689 | 687,롯,791
 690 | 688,롭,788
 691 | 689,젊,781
 692 | 690,넓,778
 693 | 691,멘,777
 694 | 692,냉,772
 695 | 693,칼,771
 696 | 694,잉,768
 697 | 695,빙,768
 698 | 696,뿐,767
 699 | 697,옮,761
 700 | 698,젤,760
 701 | 699,B,757
 702 | 700,죄,756
 703 | 701,탔,752
 704 | 702,샤,746
 705 | 703,홀,745
 706 | 704,떼,743
 707 | 705,줌,738
 708 | 706,징,734
 709 | 707,폭,727
 710 | 708,G,721
 711 | 709,킬,713
 712 | 710,흔,712
 713 | 711,딜,711
 714 | 712,슈,703
 715 | 713,율,700
 716 | 714,즌,697
 717 | 715,씀,694
 718 | 716,앙,689
 719 | 717,눠,688
 720 | 718,콩,686
 721 | 719,얻,684
 722 | 720,숨,682
 723 | 721,닝,673
 724 | 722,꽃,668
 725 | 723,쌀,667
 726 | 724,컬,666
 727 | 725,춤,666
 728 | 726,c,664
 729 | 727,뚫,663
 730 | 728,엠,661
 731 | 729,몬,659
 732 | 730,D,658
 733 | 731,흐,656
 734 | 732,앤,646
 735 | 733,똥,645
 736 | 734,콜,644
 737 | 735,델,635
 738 | 736,렀,628
 739 | 737,폐,627
 740 | 738,엘,624
 741 | 739,쁠,623
 742 | 740,랄,622
 743 | 741,걘,621
 744 | 742,벤,619
 745 | 743,봄,611
 746 | 744,왠,609
 747 | 745,씻,609
 748 | 746,률,608
 749 | 747,켰,600
 750 | 748,짱,600
 751 | 749,웹,599
 752 | 750,압,599
 753 | 751,럭,596
 754 | 752,땅,596
 755 | 753,멍,595
 756 | 754,랩,595
 757 | 755,댓,595
 758 | 756,깊,595
 759 | 757,뮤,592
 760 | 758,령,590
 761 | 759,릿,589
 762 | 760,낀,589
 763 | 761,윤,586
 764 | 762,옥,584
 765 | 763,룸,582
 766 | 764,딘,579
 767 | 765,객,578
 768 | 766,댄,576
 769 | 767,컵,574
 770 | 768,폴,573
 771 | 769,쟤,570
 772 | 770,뷰,569
 773 | 771,템,568
 774 | 772,덴,567
 775 | 773,눌,559
 776 | 774,캠,558
 777 | 775,홈,557
 778 | 776,삶,557
 779 | 777,삭,555
 780 | 778,벨,555
 781 | 779,엉,552
 782 | 780,헐,549
 783 | 781,벅,545
 784 | 782,벗,544
 785 | 783,혈,543
 786 | 784,밍,539
 787 | 785,셀,536
 788 | 786,낭,534
 789 | 787,춥,533
 790 | 788,릉,533
 791 | 789,t,533
 792 | 790,잃,532
 793 | 791,I,529
 794 | 792,놈,526
 795 | 793,춘,524
 796 | 794,찜,520
 797 | 795,R,519
 798 | 796,걷,518
 799 | 797,삐,515
 800 | 798,헌,510
 801 | 799,딨,510
 802 | 800,빛,504
 803 | 801,흘,503
 804 | 802,닫,502
 805 | 803,균,502
 806 | 804,p,495
 807 | 805,L,494
 808 | 806,좌,492
 809 | 807,껄,491
 810 | 808,펜,489
 811 | 809,N,487
 812 | 810,싹,486
 813 | 811,탑,485
 814 | 812,쏘,483
 815 | 813,O,482
 816 | 814,픽,480
 817 | 815,덩,477
 818 | 816,햄,476
 819 | 817,큐,473
 820 | 818,힐,472
 821 | 819,곧,471
 822 | 820,낳,470
 823 | 821,힌,468
 824 | 822,팩,468
 825 | 823,뒷,468
 826 | 824,툰,467
 827 | 825,섬,465
 828 | 826,꽂,463
 829 | 827,례,462
 830 | 828,핫,460
 831 | 829,섞,460
 832 | 830,촬,458
 833 | 831,흰,457
 834 | 832,둥,455
 835 | 833,K,450
 836 | 834,괴,449
 837 | 835,s,448
 838 | 836,핀,446
 839 | 837,꿨,444
 840 | 838,틱,441
 841 | 839,밝,441
 842 | 840,랙,440
 843 | 841,땠,440
 844 | 842,둔,440
 845 | 843,슴,439
 846 | 844,첨,438
 847 | 845,밴,432
 848 | 846,렁,431
 849 | 847,칭,429
 850 | 848,묻,428
 851 | 849,뜬,425
 852 | 850,깎,424
 853 | 851,엇,423
 854 | 852,컸,421
 855 | 853,퀴,420
 856 | 854,납,418
 857 | 855,협,417
 858 | 856,몽,416
 859 | 857,꼐,415
 860 | 858,떴,414
 861 | 859,썰,410
 862 | 860,찐,407
 863 | 861,꼴,407
 864 | 862,갠,406
 865 | 863,턱,405
 866 | 864,틴,398
 867 | 865,낄,398
 868 | 866,뒀,397
 869 | 867,끗,396
 870 | 868,꼼,395
 871 | 869,F,395
 872 | 870,샵,394
 873 | 871,휘,392
 874 | 872,뼈,390
 875 | 873,뚜,389
 876 | 874,쩍,388
 877 | 875,팡,386
 878 | 876,멜,386
 879 | 877,톤,385
 880 | 878,앨,385
 881 | 879,탐,384
 882 | 880,칸,384
 883 | 881,끓,383
 884 | 882,뚱,381
 885 | 883,닮,378
 886 | 884,깃,375
 887 | 885,짬,374
 888 | 886,빤,371
 889 | 887,측,370
 890 | 888,혔,369
 891 | 889,꽁,369
 892 | 890,펴,368
 893 | 891,앴,368
 894 | 892,겸,368
 895 | 893,쿨,367
 896 | 894,릇,363
 897 | 895,얀,362
 898 | 896,쿄,358
 899 | 897,컷,358
 900 | 898,팠,356
 901 | 899,끈,355
 902 | 900,렴,354
 903 | 901,잊,352
 904 | 902,덤,350
 905 | 903,갤,342
 906 | 904,븐,340
 907 | 905,흡,337
 908 | 906,덮,337
 909 | 907,씹,335
 910 | 908,뽀,335
 911 | 909,뚝,335
 912 | 910,갚,335
 913 | 911,찔,334
 914 | 912,댔,333
 915 | 913,혁,332
 916 | 914,띠,328
 917 | 915,벼,327
 918 | 916,얇,324
 919 | 917,뺐,324
 920 | 918,팝,323
 921 | 919,잇,322
 922 | 920,왼,322
 923 | 921,낚,321
 924 | 922,칙,316
 925 | 923,겉,316
 926 | 924,뜯,313
 927 | 925,닦,312
 928 | 926,짠,311
 929 | 927,썹,310
 930 | 928,뷔,310
 931 | 929,묶,310
 932 | 930,꾼,306
 933 | 931,빅,305
 934 | 932,땄,305
 935 | 933,캡,304
 936 | 934,묘,304
 937 | 935,샘,303
 938 | 936,묵,303
 939 | 937,a,302
 940 | 938,쭈,300
 941 | 939,b,300
 942 | 940,겪,299
 943 | 941,둬,298
 944 | 942,J,298
 945 | 943,쫄,296
 946 | 944,랫,296
 947 | 945,뀐,296
 948 | 946,흑,295
 949 | 947,댕,295
 950 | 948,꽉,295
 951 | 949,곰,294
 952 | 950,붕,293
 953 | 951,땀,292
 954 | 952,릎,290
 955 | 953,뽕,289
 956 | 954,쥐,288
 957 | 955,렉,287
 958 | 956,숭,283
 959 | 957,샐,283
 960 | 958,v,282
 961 | 959,렛,281
 962 | 960,녕,281
 963 | 961,힙,280
 964 | 962,쫙,279
 965 | 963,촉,278
 966 | 964,쩜,277
 967 | 965,긋,277
 968 | 966,샌,276
 969 | 967,o,275
 970 | 968,쫓,273
 971 | 969,쩐,273
 972 | 970,헷,272
 973 | 971,X,268
 974 | 972,웅,267
 975 | 973,뺏,267
 976 | 974,쵸,266
 977 | 975,쪘,266
 978 | 976,랍,266
 979 | 977,E,266
 980 | 978,좁,265
 981 | 979,앱,265
 982 | 980,썸,264
 983 | 981,냅,264
 984 | 982,펙,263
 985 | 983,늙,263
 986 | 984,껌,261
 987 | 985,n,261
 988 | 986,e,261
 989 | 987,랭,260
 990 | 988,귤,260
 991 | 989,찢,259
 992 | 990,닿,259
 993 | 991,띄,258
 994 | 992,긁,255
 995 | 993,귄,253
 996 | 994,굽,253
 997 | 995,갓,253
 998 | 996,캔,252
 999 | 997,멈,252
1000 | 998,욱,250
1001 | 999,뺄,250
1002 | 1000,뇌,250
1003 | 1001,팟,249
1004 | 1002,쌌,248
1005 | 1003,룹,248
1006 | 1004,덥,248
1007 | 1005,폼,246
1008 | 1006,톱,244
1009 | 1007,듬,244
1010 | 1008,껍,244
1011 | 1009,흠,243
1012 | 1010,팍,243
1013 | 1011,맹,243
1014 | 1012,쉴,242
1015 | 1013,썩,240
1016 | 1014,밟,240
1017 | 1015,맵,237
1018 | 1016,돋,236
1019 | 1017,콤,235
1020 | 1018,맙,234
1021 | 1019,뱅,233
1022 | 1020,쫍,231
1023 | 1021,윗,229
1024 | 1022,뜩,229
1025 | 1023,찝,228
1026 | 1024,뺀,227
1027 | 1025,닷,226
1028 | 1026,넨,226
1029 | 1027,쌈,225
1030 | 1028,쩨,224
1031 | 1029,붓,224
1032 | 1030,쩡,223
1033 | 1031,믹,223
1034 | 1032,잼,221
1035 | 1033,r,221
1036 | 1034,쭐,220
1037 | 1035,엊,219
1038 | 1036,g,219
1039 | 1037,췄,217
1040 | 1038,룩,217
1041 | 1039,텀,215
1042 | 1040,쇠,213
1043 | 1041,숫,212
1044 | 1042,풋,210
1045 | 1043,쌩,208
1046 | 1044,쾌,207
1047 | 1045,볍,207
1048 | 1046,뤄,207
1049 | 1047,겐,207
1050 | 1048,m,207
1051 | 1049,펌,206
1052 | 1050,쪄,206
1053 | 1051,뻥,206
1054 | 1052,i,206
1055 | 1053,뻤,205
1056 | 1054,k,204
1057 | 1055,핵,203
1058 | 1056,셉,200
1059 | 1057,듀,198
1060 | 1058,닉,198
1061 | 1059,략,197
1062 | 1060,넉,197
1063 | 1061,딪,196
1064 | 1062,낯,195
1065 | 1063,텍,194
1066 | 1064,뱃,194
1067 | 1065,멤,194
1068 | 1066,윈,192
1069 | 1067,엎,192
1070 | 1068,뭉,192
1071 | 1069,젝,191
1072 | 1070,셜,190
1073 | 1071,빴,190
1074 | 1072,룰,190
1075 | 1073,앗,189
1076 | 1074,궈,189
1077 | 1075,윙,188
1078 | 1076,엥,188
1079 | 1077,d,185
1080 | 1078,꼽,183
1081 | 1079,챔,182
1082 | 1080,쉐,182
1083 | 1081,봇,182
1084 | 1082,푼,180
1085 | 1083,댁,179
1086 | 1084,칵,178
1087 | 1085,뿔,177
1088 | 1086,뺑,176
1089 | 1087,탱,175
1090 | 1088,쿼,174
1091 | 1089,갱,174
1092 | 1090,퉁,173
1093 | 1091,빔,173
1094 | 1092,썬,172
1095 | 1093,빽,172
1096 | 1094,둑,172
1097 | 1095,헛,169
1098 | 1096,빗,168
1099 | 1097,탓,167
1100 | 1098,륵,165
1101 | 1099,꼰,165
1102 | 1100,쎈,163
1103 | 1101,쥬,162
1104 | 1102,깡,162
1105 | 1103,퀄,161
1106 | 1104,빚,161
1107 | 1105,즉,160
1108 | 1106,삿,160
1109 | 1107,밭,160
1110 | 1108,u,160
1111 | 1109,혐,159
1112 | 1110,햇,159
1113 | 1111,툭,159
1114 | 1112,탠,158
1115 | 1113,샷,158
1116 | 1114,맣,157
1117 | 1115,껏,157
1118 | 1116,핏,156
1119 | 1117,앵,155
1120 | 1118,뜰,155
1121 | 1119,굿,155
1122 | 1120,U,155
1123 | 1121,섹,154
1124 | 1122,펑,153
1125 | 1123,맻,153
1126 | 1124,뀔,153
1127 | 1125,깥,152
1128 | 1126,뱀,150
1129 | 1127,뢰,150
1130 | 1128,껀,150
1131 | 1129,뉘,149
1132 | 1130,흉,146
1133 | 1131,틈,146
1134 | 1132,쏟,146
1135 | 1133,훔,143
1136 | 1134,쇄,142
1137 | 1135,뎅,142
1138 | 1136,칩,141
1139 | 1137,띵,140
1140 | 1138,푹,139
1141 | 1139,넥,139
1142 | 1140,퀘,137
1143 | 1141,훅,136
1144 | 1142,융,136
1145 | 1143,멸,136
1146 | 1144,냠,136
1147 | 1145,횟,134
1148 | 1146,찼,134
1149 | 1147,Y,134
1150 | 1148,룬,133
1151 | 1149,귈,133
1152 | 1150,H,133
1153 | 1151,젓,132
1154 | 1152,쏠,132
1155 | 1153,숲,132
1156 | 1154,냬,132
1157 | 1155,l,132
1158 | 1156,짰,130
1159 | 1157,멕,130
1160 | 1158,뇨,130
1161 | 1159,팽,128
1162 | 1160,깼,127
1163 | 1161,숏,126
1164 | 1162,굔,125
1165 | 1163,슐,123
1166 | 1164,쉰,123
1167 | 1165,얄,122
1168 | 1166,뱉,122
1169 | 1167,렬,122
1170 | 1168,굶,122
1171 | 1169,팁,121
1172 | 1170,츄,121
1173 | 1171,뭣,121
1174 | 1172,륙,121
1175 | 1173,횡,120
1176 | 1174,옹,119
1177 | 1175,뻘,117
1178 | 1176,옵,116
1179 | 1177,옴,116
1180 | 1178,얹,116
1181 | 1179,쑥,116
1182 | 1180,깄,116
1183 | 1181,므,115
1184 | 1182,찡,112
1185 | 1183,젖,112
1186 | 1184,꽈,112
1187 | 1185,틸,111
1188 | 1186,콕,111
1189 | 1187,첩,111
1190 | 1188,똘,111
1191 | 1189,쿵,110
1192 | 1190,왤,110
1193 | 1191,괌,110
1194 | 1192,밸,109
1195 | 1193,녜,109
1196 | 1194,갸,109
1197 | 1195,펀,108
1198 | 1196,칫,108
1199 | 1197,맺,108
1200 | 1198,탭,107
1201 | 1199,쁨,106
1202 | 1200,폈,105
1203 | 1201,펼,105
1204 | 1202,첼,105
1205 | 1203,숱,105
1206 | 1204,섰,105
1207 | 1205,킥,104
1208 | 1206,맑,103
1209 | 1207,랗,103
1210 | 1208,펐,102
1211 | 1209,넛,102
1212 | 1210,솜,101
1213 | 1211,벙,101
1214 | 1212,껑,101
1215 | 1213,f,101
1216 | 1214,룡,100
1217 | 1215,훌,99
1218 | 1216,x,98
1219 | 1217,쓱,97
1220 | 1218,늬,97
1221 | 1219,곽,97
1222 | 1220,y,97
1223 | 1221,욘,96
1224 | 1222,돔,96
1225 | 1223,겄,96
1226 | 1224,텝,95
1227 | 1225,훠,94
1228 | 1226,텅,94
1229 | 1227,씌,94
1230 | 1228,꺾,94
1231 | 1229,벚,93
1232 | 1230,렷,93
1233 | 1231,귓,93
1234 | 1232,찹,92
1235 | 1233,툴,91
1236 | 1234,깅,91
1237 | 1235,쭤,90
1238 | 1236,욜,90
1239 | 1237,얌,90
1240 | 1238,짖,89
1241 | 1239,옳,89
1242 | 1240,벳,88
1243 | 1241,뛸,88
1244 | 1242,깠,88
1245 | 1243,퍽,87
1246 | 1244,퀸,87
1247 | 1245,엮,87
1248 | 1246,삽,87
1249 | 1247,겟,87
1250 | 1248,왓,86
1251 | 1249,댈,86
1252 | 1250,샴,85
1253 | 1251,뻗,85
1254 | 1252,됨,84
1255 | 1253,얜,83
1256 | 1254,굵,83
1257 | 1255,눕,82
1258 | 1256,갇,82
1259 | 1257,셰,81
1260 | 1258,늫,81
1261 | 1259,텨,80
1262 | 1260,숍,80
1263 | 1261,뻑,80
1264 | 1262,됩,80
1265 | 1263,잎,79
1266 | 1264,뭇,79
1267 | 1265,퐁,78
1268 | 1266,팸,78
1269 | 1267,쯔,78
1270 | 1268,넜,78
1271 | 1269,깍,78
1272 | 1270,쌔,77
1273 | 1271,셈,77
1274 | 1272,읍,76
1275 | 1273,픔,75
1276 | 1274,펫,75
1277 | 1275,콧,75
1278 | 1276,얗,75
1279 | 1277,눅,75
1280 | 1278,j,74
1281 | 1279,쬐,73
1282 | 1280,렙,73
1283 | 1281,닙,73
1284 | 1282,슥,72
1285 | 1283,흙,71
1286 | 1284,쭝,71
1287 | 1285,짭,71
1288 | 1286,샹,71
1289 | 1287,릏,71
1290 | 1288,럿,71
1291 | 1289,덧,71
1292 | 1290,즙,70
1293 | 1291,늑,70
1294 | 1292,괄,70
1295 | 1293,킷,68
1296 | 1294,쿡,68
1297 | 1295,캉,68
1298 | 1296,둡,68
1299 | 1297,톨,67
1300 | 1298,엣,67
1301 | 1299,숟,67
1302 | 1300,낑,67
1303 | 1301,펭,66
1304 | 1302,왁,66
1305 | 1303,쏴,66
1306 | 1304,쏙,66
1307 | 1305,봅,66
1308 | 1306,멧,66
1309 | 1307,줏,65
1310 | 1308,뵈,65
1311 | 1309,쫑,64
1312 | 1310,륨,64
1313 | 1311,h,64
1314 | 1312,펄,63
1315 | 1313,짼,63
1316 | 1314,짚,63
1317 | 1315,껐,63
1318 | 1316,겜,63
1319 | 1317,싯,62
1320 | 1318,붐,62
1321 | 1319,렐,62
1322 | 1320,돗,62
1323 | 1321,팥,61
1324 | 1322,웰,61
1325 | 1323,륜,61
1326 | 1324,잣,60
1327 | 1325,슝,60
1328 | 1326,붉,60
1329 | 1327,윽,59
1330 | 1328,삘,59
1331 | 1329,딲,59
1332 | 1330,갯,58
1333 | 1331,횐,57
1334 | 1332,헨,57
1335 | 1333,캘,57
1336 | 1334,쩰,57
1337 | 1335,뤘,57
1338 | 1336,랴,57
1339 | 1337,껜,57
1340 | 1338,펠,56
1341 | 1339,킵,56
1342 | 1340,컹,56
1343 | 1341,렘,56
1344 | 1342,뛴,56
1345 | 1343,헝,55
1346 | 1344,씽,55
1347 | 1345,뮬,55
1348 | 1346,젯,54
1349 | 1347,샜,54
1350 | 1348,뿜,54
1351 | 1349,뒹,54
1352 | 1350,뎌,54
1353 | 1351,깬,54
1354 | 1352,챠,53
1355 | 1353,왈,53
1356 | 1354,뾰,53
1357 | 1355,뚤,53
1358 | 1356,꾹,53
1359 | 1357,갛,53
1360 | 1358,잌,52
1361 | 1359,엿,52
1362 | 1360,솥,52
1363 | 1361,벡,52
1364 | 1362,룻,52
1365 | 1363,꿍,52
1366 | 1364,곈,52
1367 | 1365,팜,51
1368 | 1366,튕,51
1369 | 1367,컥,51
1370 | 1368,첸,51
1371 | 1369,줍,51
1372 | 1370,섀,51
1373 | 1371,몫,51
1374 | 1372,뜸,51
1375 | 1373,깁,51
1376 | 1374,핬,50
1377 | 1375,쭘,50
1378 | 1376,쌰,50
1379 | 1377,넬,50
1380 | 1378,큘,49
1381 | 1379,쾅,49
1382 | 1380,캄,48
1383 | 1381,괘,48
1384 | 1382,쟀,47
1385 | 1383,윌,47
1386 | 1384,엌,47
1387 | 1385,앓,47
1388 | 1386,씁,47
1389 | 1387,륭,47
1390 | 1388,W,47
1391 | 1389,쑤,46
1392 | 1390,삥,46
1393 | 1391,돕,46
1394 | 1392,깰,46
1395 | 1393,핍,45
1396 | 1394,텃,45
1397 | 1395,슛,45
1398 | 1396,맸,45
1399 | 1397,롬,45
1400 | 1398,갭,45
1401 | 1399,얽,44
1402 | 1400,쏭,44
1403 | 1401,랠,44
1404 | 1402,겔,44
1405 | 1403,Q,44
1406 | 1404,핥,43
1407 | 1405,킴,43
1408 | 1406,읏,43
1409 | 1407,앚,43
1410 | 1408,숯,43
1411 | 1409,밋,43
1412 | 1410,뽐,42
1413 | 1411,뻣,42
1414 | 1412,눴,42
1415 | 1413,잭,41
1416 | 1414,뽈,41
1417 | 1415,뗐,41
1418 | 1416,꽝,41
1419 | 1417,훑,40
1420 | 1418,캥,40
1421 | 1419,쫀,40
1422 | 1420,뵙,40
1423 | 1421,홋,39
1424 | 1422,펍,39
1425 | 1423,뺨,39
1426 | 1424,됬,39
1427 | 1425,끽,39
1428 | 1426,빕,38
1429 | 1427,밉,38
1430 | 1428,꿋,38
1431 | 1429,헉,37
1432 | 1430,캣,37
1433 | 1431,촘,37
1434 | 1432,셌,37
1435 | 1433,삔,37
1436 | 1434,삑,37
1437 | 1435,뽂,37
1438 | 1436,뮌,37
1439 | 1437,뗄,37
1440 | 1438,텼,36
1441 | 1439,탬,36
1442 | 1440,쨍,36
1443 | 1441,웜,36
1444 | 1442,앰,36
1445 | 1443,맴,36
1446 | 1444,띡,36
1447 | 1445,꿇,36
1448 | 1446,걀,36
1449 | 1447,흩,35
1450 | 1448,쥴,35
1451 | 1449,씸,35
1452 | 1450,낡,35
1453 | 1451,곁,35
1454 | 1452,휙,34
1455 | 1453,쿤,34
1456 | 1454,켄,34
1457 | 1455,츤,34
1458 | 1456,얕,34
1459 | 1457,썪,34
1460 | 1458,둠,34
1461 | 1459,촛,33
1462 | 1460,챌,33
1463 | 1461,죙,33
1464 | 1462,쟈,33
1465 | 1463,잰,33
1466 | 1464,뵀,33
1467 | 1465,w,33
1468 | 1466,푠,32
1469 | 1467,폿,32
1470 | 1468,쨈,32
1471 | 1469,밲,32
1472 | 1470,랖,32
1473 | 1471,떵,32
1474 | 1472,찻,31
1475 | 1473,옇,31
1476 | 1474,뽁,31
1477 | 1475,둣,31
1478 | 1476,닳,31
1479 | 1477,긱,31
1480 | 1478,곶,31
1481 | 1479,휠,30
1482 | 1480,춧,30
1483 | 1481,쐈,30
1484 | 1482,썽,30
1485 | 1483,뼛,29
1486 | 1484,떳,29
1487 | 1485,굘,29
1488 | 1486,훗,28
1489 | 1487,퀵,28
1490 | 1488,봬,28
1491 | 1489,릅,28
1492 | 1490,꺠,28
1493 | 1491,슉,27
1494 | 1492,눔,27
1495 | 1493,끙,27
1496 | 1494,궐,27
1497 | 1495,2,27
1498 | 1496,촥,26
1499 | 1497,젬,26
1500 | 1498,솟,26
1501 | 1499,맷,26
1502 | 1500,룐,26
1503 | 1501,뎃,26
1504 | 1502,깽,26
1505 | 1503,툼,25
1506 | 1504,쎌,25
1507 | 1505,쉼,25
1508 | 1506,쉘,25
1509 | 1507,숑,25
1510 | 1508,뎀,25
1511 | 1509,냔,25
1512 | 1510,쫘,24
1513 | 1511,쎘,24
1514 | 1512,싣,24
1515 | 1513,섣,24
1516 | 1514,샾,24
1517 | 1515,맬,24
1518 | 1516,뗀,24
1519 | 1517,꾀,24
1520 | 1518,헹,23
1521 | 1519,햐,23
1522 | 1520,톰,23
1523 | 1521,췌,23
1524 | 1522,챈,23
1525 | 1523,봔,23
1526 | 1524,밧,23
1527 | 1525,맏,23
1528 | 1526,딥,23
1529 | 1527,늠,23
1530 | 1528,낵,23
1531 | 1529,낱,23
1532 | 1530,꺄,23
1533 | 1531,갬,23
1534 | 1532,훼,22
1535 | 1533,핼,22
1536 | 1534,튠,22
1537 | 1535,웩,22
1538 | 1536,쏜,22
1539 | 1537,뿅,22
1540 | 1538,빰,22
1541 | 1539,딤,22
1542 | 1540,꿉,22
1543 | 1541,걜,22
1544 | 1542,1,22
1545 | 1543,짙,21
1546 | 1544,얍,21
1547 | 1545,샛,21
1548 | 1546,뗘,21
1549 | 1547,듭,21
1550 | 1548,챘,20
1551 | 1549,쯧,20
1552 | 1550,짹,20
1553 | 1551,잦,20
1554 | 1552,옐,20
1555 | 1553,빳,20
1556 | 1554,몹,20
1557 | 1555,몄,20
1558 | 1556,똔,20
1559 | 1557,딧,20
1560 | 1558,놉,20
1561 | 1559,궜,20
1562 | 1560,굼,20
1563 | 1561,헥,19
1564 | 1562,캬,19
1565 | 1563,챕,19
1566 | 1564,쟨,19
1567 | 1565,멓,19
1568 | 1566,똠,19
1569 | 1567,댐,19
1570 | 1568,텁,18
1571 | 1569,켈,18
1572 | 1570,첵,18
1573 | 1571,숄,18
1574 | 1572,띨,18
1575 | 1573,듦,18
1576 | 1574,궤,18
1577 | 1575,곗,18
1578 | 1576,튈,17
1579 | 1577,좆,17
1580 | 1578,윷,17
1581 | 1579,옅,17
1582 | 1580,얏,17
1583 | 1581,믈,17
1584 | 1582,룽,17
1585 | 1583,띃,17
1586 | 1584,딕,17
1587 | 1585,뎁,17
1588 | 1586,닛,17
1589 | 1587,냑,17
1590 | 1588,겅,17
1591 | 1589,휩,16
1592 | 1590,팎,16
1593 | 1591,틋,16
1594 | 1592,콸,16
1595 | 1593,콥,16
1596 | 1594,잴,16
1597 | 1595,웁,16
1598 | 1596,슘,16
1599 | 1597,멱,16
1600 | 1598,랏,16
1601 | 1599,떄,16
1602 | 1600,뒨,16
1603 | 1601,꿰,16
1604 | 1602,깻,16
1605 | 1603,긌,16
1606 | 1604,젼,15
1607 | 1605,윰,15
1608 | 1606,웍,15
1609 | 1607,앳,15
1610 | 1608,샬,15
1611 | 1609,샥,15
1612 | 1610,볕,15
1613 | 1611,멩,15
1614 | 1612,넹,15
1615 | 1613,넙,15
1616 | 1614,끕,15
1617 | 1615,휜,14
1618 | 1616,텄,14
1619 | 1617,쫒,14
1620 | 1618,쩝,14
1621 | 1619,쨋,14
1622 | 1620,윳,14
1623 | 1621,쉭,14
1624 | 1622,쇽,14
1625 | 1623,셧,14
1626 | 1624,뵐,14
1627 | 1625,땔,14
1628 | 1626,덱,14
1629 | 1627,댑,14
1630 | 1628,꺽,14
1631 | 1629,곪,14
1632 | 1630,켔,13
1633 | 1631,츰,13
1634 | 1632,읜,13
1635 | 1633,쑈,13
1636 | 1634,볐,13
1637 | 1635,및,13
1638 | 1636,롷,13
1639 | 1637,딛,13
1640 | 1638,냇,13
1641 | 1639,3,13
1642 | 1640,혓,12
1643 | 1641,팻,12
1644 | 1642,팰,12
1645 | 1643,킁,12
1646 | 1644,촤,12
1647 | 1645,쨰,12
1648 | 1646,잿,12
1649 | 1647,옌,12
1650 | 1648,쐬,12
1651 | 1649,쌋,12
1652 | 1650,슁,12
1653 | 1651,쉑,12
1654 | 1652,빢,12
1655 | 1653,뵌,12
1656 | 1654,뭄,12
1657 | 1655,묽,12
1658 | 1656,뎠,12
1659 | 1657,늪,12
1660 | 1658,뇽,12
1661 | 1659,훤,11
1662 | 1660,횔,11
1663 | 1661,홧,11
1664 | 1662,햅,11
1665 | 1663,푯,11
1666 | 1664,팼,11
1667 | 1665,탯,11
1668 | 1666,탤,11
1669 | 1667,큭,11
1670 | 1668,짢,11
1671 | 1669,읊,11
1672 | 1670,롸,11
1673 | 1671,띤,11
1674 | 1672,놋,11
1675 | 1673,넴,11
1676 | 1674,귿,11
1677 | 1675,q,11
1678 | 1676,휑,10
1679 | 1677,퓸,10
1680 | 1678,튿,10
1681 | 1679,튄,10
1682 | 1680,촐,10
1683 | 1681,쭌,10
1684 | 1682,짊,10
1685 | 1683,숴,10
1686 | 1684,숀,10
1687 | 1685,뿡,10
1688 | 1686,뻬,10
1689 | 1687,렜,10
1690 | 1688,뗬,10
1691 | 1689,늄,10
1692 | 1690,끍,10
1693 | 1691,곌,10
1694 | 1692,갰,10
1695 | 1693,폄,9
1696 | 1694,콰,9
1697 | 1695,캇,9
1698 | 1696,캅,9
1699 | 1697,늉,9
1700 | 1698,뉜,9
1701 | 1699,큔,8
1702 | 1700,콱,8
1703 | 1701,켠,8
1704 | 1702,쳔,8
1705 | 1703,쳇,8
1706 | 1704,읔,8
1707 | 1705,읎,8
1708 | 1706,욤,8
1709 | 1707,뼘,8
1710 | 1708,롹,8
1711 | 1709,렝,8
1712 | 1710,뚠,8
1713 | 1711,땋,8
1714 | 1712,덨,8
1715 | 1713,넵,8
1716 | 1714,넝,8
1717 | 1715,넋,8
1718 | 1716,꿩,8
1719 | 1717,꼿,8
1720 | 1718,깟,8
1721 | 1719,곯,8
1722 | 1720,4,8
1723 | 1721,힝,7
1724 | 1722,헀,7
1725 | 1723,푤,7
1726 | 1724,쿱,7
1727 | 1725,캤,7
1728 | 1726,챗,7
1729 | 1727,쯩,7
1730 | 1728,쮸,7
1731 | 1729,읗,7
1732 | 1730,윅,7
1733 | 1731,얠,7
1734 | 1732,씰,7
1735 | 1733,썅,7
1736 | 1734,쌜,7
1737 | 1735,쌉,7
1738 | 1736,슌,7
1739 | 1737,쉿,7
1740 | 1738,쇳,7
1741 | 1739,셍,7
1742 | 1740,뿍,7
1743 | 1741,뼌,7
1744 | 1742,뺌,7
1745 | 1743,밈,7
1746 | 1744,룟,7
1747 | 1745,룔,7
1748 | 1746,뜀,7
1749 | 1747,끅,7
1750 | 1748,꾜,7
1751 | 1749,겡,7
1752 | 1750,Z,7
1753 | 1751,횰,6
1754 | 1752,헴,6
1755 | 1753,핌,6
1756 | 1754,펩,6
1757 | 1755,펨,6
1758 | 1756,켸,6
1759 | 1757,쩠,6
1760 | 1758,잽,6
1761 | 1759,엡,6
1762 | 1760,앝,6
1763 | 1761,쓕,6
1764 | 1762,썜,6
1765 | 1763,쌘,6
1766 | 1764,삣,6
1767 | 1765,빻,6
1768 | 1766,몀,6
1769 | 1767,뤼,6
1770 | 1768,롄,6
1771 | 1769,떫,6
1772 | 1770,덫,6
1773 | 1771,뉸,6
1774 | 1772,꽥,6
1775 | 1773,궂,6
1776 | 1774,괭,6
1777 | 1775,0,6
1778 | 1776,힉,5
1779 | 1777,휸,5
1780 | 1778,퓰,5
1781 | 1779,퉤,5
1782 | 1780,퀭,5
1783 | 1781,켐,5
1784 | 1782,췻,5
1785 | 1783,챡,5
1786 | 1784,쨀,5
1787 | 1785,젱,5
1788 | 1786,읒,5
1789 | 1787,웡,5
1790 | 1788,옙,5
1791 | 1789,얬,5
1792 | 1790,앎,5
1793 | 1791,씼,5
1794 | 1792,쑨,5
1795 | 1793,쐐,5
1796 | 1794,쌨,5
1797 | 1795,쉈,5
1798 | 1796,숩,5
1799 | 1797,셤,5
1800 | 1798,삠,5
1801 | 1799,뱄,5
1802 | 1800,뭥,5
1803 | 1801,멨,5
1804 | 1802,먀,5
1805 | 1803,랒,5
1806 | 1804,땟,5
1807 | 1805,땁,5
1808 | 1806,뉠,5
1809 | 1807,꽌,5
1810 | 1808,귐,5
1811 | 1809,굥,5
1812 | 1810,굣,5
1813 | 1811,겋,5
1814 | 1812,8,5
1815 | 1813,5,5
1816 | 1814,훙,4
1817 | 1815,혠,4
1818 | 1816,폔,4
1819 | 1817,튬,4
1820 | 1818,퉜,4
1821 | 1819,탉,4
1822 | 1820,큽,4
1823 | 1821,퀀,4
1824 | 1822,쿰,4
1825 | 1823,켤,4
1826 | 1824,켕,4
1827 | 1825,칡,4
1828 | 1826,쯘,4
1829 | 1827,짯,4
1830 | 1828,짔,4
1831 | 1829,쥔,4
1832 | 1830,줜,4
1833 | 1831,죗,4
1834 | 1832,욧,4
1835 | 1833,쒸,4
1836 | 1834,쏼,4
1837 | 1835,쌂,4
1838 | 1836,셴,4
1839 | 1837,샨,4
1840 | 1838,뽜,4
1841 | 1839,뻠,4
1842 | 1840,빝,4
1843 | 1841,봥,4
1844 | 1842,뫼,4
1845 | 1843,맜,4
1846 | 1844,맀,4
1847 | 1845,뤠,4
1848 | 1846,띌,4
1849 | 1847,띈,4
1850 | 1848,뛌,4
1851 | 1849,뚸,4
1852 | 1850,떽,4
1853 | 1851,떰,4
1854 | 1852,떈,4
1855 | 1853,땍,4
1856 | 1854,돠,4
1857 | 1855,뎄,4
1858 | 1856,늣,4
1859 | 1857,뇸,4
1860 | 1858,놥,4
1861 | 1859,녈,4
1862 | 1860,넒,4
1863 | 1861,낏,4
1864 | 1862,갉,4
1865 | 1863,z,4
1866 | 1864,힛,3
1867 | 1865,휀,3
1868 | 1866,헙,3
1869 | 1867,픗,3
1870 | 1868,퐈,3
1871 | 1869,팹,3
1872 | 1870,틔,3
1873 | 1871,퇸,3
1874 | 1872,텟,3
1875 | 1873,탰,3
1876 | 1874,퀼,3
1877 | 1875,칬,3
1878 | 1876,췰,3
1879 | 1877,쳥,3
1880 | 1878,챂,3
1881 | 1879,찧,3
1882 | 1880,쭙,3
1883 | 1881,쬬,3
1884 | 1882,쩄,3
1885 | 1883,쨉,3
1886 | 1884,줴,3
1887 | 1885,죵,3
1888 | 1886,좍,3
1889 | 1887,좃,3
1890 | 1888,읐,3
1891 | 1889,읃,3
1892 | 1890,웟,3
1893 | 1891,왯,3
1894 | 1892,옭,3
1895 | 1893,씅,3
1896 | 1894,쒀,3
1897 | 1895,쑹,3
1898 | 1896,쏵,3
1899 | 1897,쎔,3
1900 | 1898,쎅,3
1901 | 1899,슾,3
1902 | 1900,쉣,3
1903 | 1901,솨,3
1904 | 1902,셸,3
1905 | 1903,셩,3
1906 | 1904,삯,3
1907 | 1905,뿟,3
1908 | 1906,뾱,3
1909 | 1907,뽝,3
1910 | 1908,뻰,3
1911 | 1909,뵤,3
1912 | 1910,볏,3
1913 | 1911,뱌,3
1914 | 1912,뱁,3
1915 | 1913,밨,3
1916 | 1914,밎,3
1917 | 1915,믐,3
1918 | 1916,뭡,3
1919 | 1917,뭠,3
1920 | 1918,뭍,3
1921 | 1919,묭,3
1922 | 1920,뫠,3
1923 | 1921,멏,3
1924 | 1922,멎,3
1925 | 1923,맽,3
1926 | 1924,뙤,3
1927 | 1925,떔,3
1928 | 1926,땃,3
1929 | 1927,덷,3
1930 | 1928,댜,3
1931 | 1929,늰,3
1932 | 1930,놘,3
1933 | 1931,냘,3
1934 | 1932,뀜,3
1935 | 1933,꽹,3
1936 | 1934,꽐,3
1937 | 1935,꼇,3
1938 | 1936,껸,3
1939 | 1937,껨,3
1940 | 1938,꺅,3
1941 | 1939,곘,3
1942 | 1940,겝,3
1943 | 1941,흣,2
1944 | 1942,흝,2
1945 | 1943,흄,2
1946 | 1944,헸,2
1947 | 1945,헜,2
1948 | 1946,햘,2
1949 | 1947,햑,2
1950 | 1948,핟,2
1951 | 1949,픙,2
1952 | 1950,픕,2
1953 | 1951,푀,2
1954 | 1952,퐉,2
1955 | 1953,틑,2
1956 | 1954,틍,2
1957 | 1955,큣,2
1958 | 1956,퀏,2
1959 | 1957,콴,2
1960 | 1958,켁,2
1961 | 1959,칟,2
1962 | 1960,츳,2
1963 | 1961,츨,2
1964 | 1962,츈,2
1965 | 1963,쵝,2
1966 | 1964,촙,2
1967 | 1965,찦,2
1968 | 1966,찟,2
1969 | 1967,쭁,2
1970 | 1968,쫏,2
1971 | 1969,쩬,2
1972 | 1970,쩧,2
1973 | 1971,짆,2
1974 | 1972,쥰,2
1975 | 1973,쥘,2
1976 | 1974,좐,2
1977 | 1975,졍,2
1978 | 1976,쟝,2
1979 | 1977,잧,2
1980 | 1978,잍,2
1981 | 1979,욍,2
1982 | 1980,왬,2
1983 | 1981,옫,2
1984 | 1982,옜,2
1985 | 1983,옉,2
1986 | 1984,엤,2
1987 | 1985,얐,2
1988 | 1986,얉,2
1989 | 1987,쒯,2
1990 | 1988,쑐,2
1991 | 1989,쏸,2
1992 | 1990,썻,2
1993 | 1991,쌕,2
1994 | 1992,싷,2
1995 | 1993,솝,2
1996 | 1994,솎,2
1997 | 1995,샅,2
1998 | 1996,삻,2
1999 | 1997,쁩,2
2000 | 1998,빘,2
2001 | 1999,뷸,2
2002 | 2000,뷴,2
2003 | 2001,봽,2
2004 | 2002,봈,2
2005 | 2003,볌,2
2006 | 2004,벋,2
2007 | 2005,믕,2
2008 | 2006,뮴,2
2009 | 2007,뭬,2
2010 | 2008,뭑,2
2011 | 2009,묜,2
2012 | 2010,맇,2
2013 | 2011,릈,2
2014 | 2012,륑,2
2015 | 2013,랮,2
2016 | 2014,랟,2
2017 | 2015,띔,2
2018 | 2016,떱,2
2019 | 2017,듈,2
2020 | 2018,됴,2
2021 | 2019,됫,2
2022 | 2020,됙,2
2023 | 2021,됑,2
2024 | 2022,댤,2
2025 | 2023,닯,2
2026 | 2024,늗,2
2027 | 2025,뉩,2
2028 | 2026,눟,2
2029 | 2027,눗,2
2030 | 2028,넚,2
2031 | 2029,냡,2
2032 | 2030,낢,2
2033 | 2031,꾿,2
2034 | 2032,꼳,2
2035 | 2033,꼄,2
2036 | 2034,겊,2
2037 | 2035,갼,2
2038 | 2036,6,2
2039 | 2037,<s>,0
2040 | 2038,</s>,0
2041 | 2039,_,0
2042 | 


--------------------------------------------------------------------------------
/main.py:
--------------------------------------------------------------------------------
  1 | import os
  2 | import time
  3 | import torch
  4 | import torch.nn as nn
  5 | import queue
  6 | import math
  7 | from random import random
  8 | from torch import optim
  9 | from package.config import Config
 10 | from package.definition import char2id, logger, SOS_token, EOS_token, PAD_token
 11 | from package.data_loader import CustomDataset, load_corpus, CustomDataLoader
 12 | from package.evaluator import evaluate
 13 | from package.loss import Perplexity
 14 | from package.trainer import supervised_train
 15 | from model import LanguageModel
 16 | 
 17 | # Character-level Recurrent Neural Network Language Model implement in Pytorch
 18 | # https://github.com/sooftware/char-rnnlm
 19 | 
 20 | if __name__ == '__main__':
 21 |     os.environ["CUDA_LAUNCH_BLOCKING"] = "1"  # if you use Multi-GPU, delete this line
 22 |     logger.info("device : %s" % torch.cuda.get_device_name(0))
 23 |     logger.info("CUDA is available : %s" % (torch.cuda.is_available()))
 24 |     logger.info("CUDA version : %s" % torch.version.cuda)
 25 |     logger.info("PyTorch version : %s" % torch.__version__)
 26 | 
 27 |     config = Config(
 28 |         use_cuda=True,
 29 |         hidden_size=512,
 30 |         dropout_p=0.5,
 31 |         n_layers=4,
 32 |         batch_size=16,
 33 |         max_epochs=40,
 34 |         lr=0.0001,
 35 |         teacher_forcing_ratio=1.0,
 36 |         seed=1,
 37 |         max_len=428,
 38 |         worker_num=1
 39 |     )
 40 | 
 41 |     torch.manual_seed(config.seed)
 42 |     torch.cuda.manual_seed_all(config.seed)
 43 |     cuda = config.use_cuda and torch.cuda.is_available()
 44 |     device = torch.device('cuda' if cuda else 'cpu')
 45 | 
 46 |     model = LanguageModel(
 47 |         n_class=len(char2id),
 48 |         n_layers=config.n_layers,
 49 |         rnn_cell='lstm',
 50 |         hidden_size=config.hidden_size,
 51 |         dropout_p=config.dropout_p,
 52 |         max_length=config.max_len,
 53 |         sos_id=SOS_token,
 54 |         eos_id=EOS_token,
 55 |         device=device
 56 |     )
 57 |     model.flatten_parameters()
 58 |     model = nn.DataParallel(model).to(device)
 59 | 
 60 |     for param in model.parameters():
 61 |         param.data.uniform_(-0.08, 0.08)
 62 | 
 63 |     # Prepare loss
 64 |     weight = torch.ones(len(char2id)).to(device)
 65 |     perplexity = Perplexity(weight, PAD_token, device)
 66 |     optimizer = optim.Adam(model.module.parameters(), lr=config.lr)
 67 | 
 68 |     corpus = load_corpus('./data/corpus_df.bin')
 69 |     total_time_step = math.ceil(len(corpus) / config.batch_size)
 70 | 
 71 |     train_set = CustomDataset(corpus[:-10000], SOS_token, EOS_token, config.batch_size)
 72 |     valid_set = CustomDataset(corpus[-10000:], SOS_token, EOS_token, config.batch_size)
 73 | 
 74 |     logger.info('start')
 75 |     train_begin = time.time()
 76 | 
 77 |     for epoch in range(config.max_epochs):
 78 |         train_queue = queue.Queue(config.worker_num << 1)
 79 |         train_set.shuffle()
 80 | 
 81 |         train_loader = CustomDataLoader(train_set, train_queue, config.batch_size, 0)
 82 |         train_loader.start()
 83 | 
 84 |         train_loss = supervised_train(
 85 |             model=model,
 86 |             queue=train_queue,
 87 |             epoch=epoch,
 88 |             total_time_step=total_time_step,
 89 |             train_begin=train_begin,
 90 |             perplexity=perplexity,
 91 |             optimizer=optimizer,
 92 |             device=device,
 93 |             print_every=10,
 94 |             teacher_forcing_ratio=config.teacher_forcing_ratio,
 95 |             worker_num=config.worker_num
 96 |         )
 97 | 
 98 |         torch.save(model, "./data/epoch%s.pt" % str(epoch))
 99 |         logger.info('Epoch %d (Training) Loss %0.4f' % (epoch, train_loss))
100 |         train_loader.join()
101 | 
102 |         valid_queue = queue.Queue(config.worker_num << 1)
103 |         valid_loader = CustomDataLoader(valid_set, valid_queue, config.batch_size, 0)
104 |         valid_loader.start()
105 | 
106 |         valid_loss = evaluate(model, valid_queue, perplexity, device)
107 |         valid_loader.join()
108 | 
109 |         logger.info('Epoch %d (Evaluate) Loss %0.4f' % (epoch, valid_loss))


--------------------------------------------------------------------------------
/model.py:
--------------------------------------------------------------------------------
 1 | import random
 2 | import torch
 3 | import torch.nn as nn
 4 | import torch.nn.functional as F
 5 | 
 6 | 
 7 | class LanguageModel(nn.Module):
 8 |     def __init__(self, n_class, n_layers, rnn_cell, hidden_size, dropout_p, max_length, sos_id, eos_id, device):
 9 | 
10 |         super(LanguageModel, self).__init__()
11 |         assert rnn_cell.lower() in ('lstm', 'gru', 'rnn')
12 | 
13 |         self.rnn_cell = nn.LSTM if rnn_cell.lower() == 'lstm' else nn.GRU if rnn_cell.lower() == 'gru' else nn.RNN
14 |         self.rnn = self.rnn_cell(hidden_size, hidden_size, n_layers, batch_first=True, dropout=dropout_p).to(device)
15 |         self.max_length = max_length
16 |         self.eos_id = eos_id
17 |         self.sos_id = sos_id
18 |         self.hidden_size = hidden_size
19 |         self.embedding = nn.Embedding(n_class, hidden_size)
20 |         self.n_layers = n_layers
21 |         self.input_dropout = nn.Dropout(p=dropout_p)
22 |         self.out = nn.Linear(self.hidden_size, n_class)
23 |         self.device = device
24 | 
25 |     def forward_step(self, input, hidden, function=F.log_softmax):
26 |         """ forward one time step """
27 |         batch_size = input.size(0)
28 |         seq_length = input.size(1)
29 | 
30 |         embedded = self.embedding(input).to(self.device)
31 |         embedded = self.input_dropout(embedded)
32 | 
33 |         if self.training:
34 |             self.rnn.flatten_parameters()
35 | 
36 |         output, hidden = self.rnn(embedded, hidden)
37 | 
38 |         predicted_softmax = function(self.out(output.contiguous().view(-1, self.hidden_size)), dim=1)
39 |         predicted_softmax = predicted_softmax.view(batch_size, seq_length, -1)
40 | 
41 |         return predicted_softmax, hidden
42 | 
43 |     def forward(self, inputs, teacher_forcing_ratio=1.0, function=F.log_softmax):
44 |         batch_size = inputs.size(0)
45 |         max_length = inputs.size(1) - 1  # minus the start of sequence symbol
46 | 
47 |         outputs = list()
48 |         use_teacher_forcing = True if random.random() < teacher_forcing_ratio else False
49 | 
50 |         hidden = self._init_state(batch_size)
51 | 
52 |         if use_teacher_forcing:
53 |             inputs = inputs[inputs != self.eos_id].view(batch_size, -1)
54 |             predicted_softmax, hidden = self.forward_step(
55 |                 input=inputs,
56 |                 hidden=hidden,
57 |                 function=function
58 |             )
59 | 
60 |             for di in range(predicted_softmax.size(1)):
61 |                 step_output = predicted_softmax[:, di, :]
62 |                 outputs.append(step_output)
63 | 
64 |         else:
65 |             input = inputs[:, 0].unsqueeze(1)
66 |             for di in range(max_length):
67 |                 predicted_softmax, hidden = self.forward_step(
68 |                     input=input,
69 |                     hidden=hidden,
70 |                     function=function
71 |                 )
72 | 
73 |                 step_output = predicted_softmax.squeeze(1)
74 |                 outputs.append(step_output)
75 |                 input = outputs[-1].topk(1)[1]
76 | 
77 |         return outputs
78 | 
79 |     def _init_state(self, batch_size):
80 |         if isinstance(self.rnn, nn.LSTM):
81 |             h_0 = torch.zeros(self.n_layers, batch_size, self.hidden_size).to(self.device)
82 |             c_0 = torch.zeros(self.n_layers, batch_size, self.hidden_size).to(self.device)
83 |             hidden = (h_0, c_0)
84 | 
85 |         else:
86 |             hidden = torch.zeros(self.n_layers, batch_size, self.hidden_size).to(self.device)
87 | 
88 |         return hidden
89 | 
90 |     def flatten_parameters(self):
91 |         self.rnn.flatten_parameters()


--------------------------------------------------------------------------------
/package/__pycache__/config.cpython-37.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/sooftware/char-rnnlm/fc6573bde13b151f373fc081f63e3f563debf56c/package/__pycache__/config.cpython-37.pyc


--------------------------------------------------------------------------------
/package/__pycache__/data_loader.cpython-37.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/sooftware/char-rnnlm/fc6573bde13b151f373fc081f63e3f563debf56c/package/__pycache__/data_loader.cpython-37.pyc


--------------------------------------------------------------------------------
/package/__pycache__/definition.cpython-37.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/sooftware/char-rnnlm/fc6573bde13b151f373fc081f63e3f563debf56c/package/__pycache__/definition.cpython-37.pyc


--------------------------------------------------------------------------------
/package/__pycache__/evaluator.cpython-37.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/sooftware/char-rnnlm/fc6573bde13b151f373fc081f63e3f563debf56c/package/__pycache__/evaluator.cpython-37.pyc


--------------------------------------------------------------------------------
/package/__pycache__/loss.cpython-37.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/sooftware/char-rnnlm/fc6573bde13b151f373fc081f63e3f563debf56c/package/__pycache__/loss.cpython-37.pyc


--------------------------------------------------------------------------------
/package/__pycache__/trainer.cpython-37.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/sooftware/char-rnnlm/fc6573bde13b151f373fc081f63e3f563debf56c/package/__pycache__/trainer.cpython-37.pyc


--------------------------------------------------------------------------------
/package/__pycache__/utils.cpython-37.pyc:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/sooftware/char-rnnlm/fc6573bde13b151f373fc081f63e3f563debf56c/package/__pycache__/utils.cpython-37.pyc


--------------------------------------------------------------------------------
/package/config.py:
--------------------------------------------------------------------------------
 1 | class Config:
 2 |     def __init__(self,
 3 |                  use_cuda=True,
 4 |                  hidden_size=512,
 5 |                  dropout_p=0.5,
 6 |                  n_layers=4,
 7 |                  batch_size=32,
 8 |                  max_epochs=40,
 9 |                  lr=0.0001,
10 |                  teacher_forcing_ratio=1.0,
11 |                  seed=1,
12 |                  max_len=428,
13 |                  worker_num=1
14 |                  ):
15 |         self.use_cuda = use_cuda
16 |         self.hidden_size = hidden_size
17 |         self.dropout_p = dropout_p
18 |         self.n_layers = n_layers
19 |         self.batch_size = batch_size
20 |         self.max_epochs = max_epochs
21 |         self.lr = lr
22 |         self.teacher_forcing_ratio = teacher_forcing_ratio
23 |         self.seed = seed
24 |         self.max_len = max_len
25 |         self.worker_num = worker_num
26 | 


--------------------------------------------------------------------------------
/package/data_loader.py:
--------------------------------------------------------------------------------
  1 | import threading
  2 | import csv
  3 | import random
  4 | import pandas as pd
  5 | import torch
  6 | import math
  7 | import pickle
  8 | from torch.utils.data import Dataset
  9 | from package.definition import logger
 10 | from package.utils import get_label, get_input
 11 | 
 12 | 
 13 | class CustomDataLoader(threading.Thread):
 14 |     def __init__(self, dataset, queue, batch_size, thread_id):
 15 |         threading.Thread.__init__(self)
 16 |         self.collate_fn = _collate_fn
 17 |         self.dataset = dataset
 18 |         self.batch_size = batch_size
 19 |         self.queue = queue
 20 |         self.index = 0
 21 |         self.thread_id = thread_id
 22 |         self.dataset_count = dataset.count()
 23 | 
 24 |     def create_empty_batch(self):
 25 |         sequences = torch.zeros(0, 0, 0).to(torch.long)
 26 |         targets = torch.zeros(0, 0, 0).to(torch.long)
 27 | 
 28 |         sequence_lengths = list()
 29 |         target_lengths = list()
 30 | 
 31 |         return sequences, targets, sequence_lengths, target_lengths
 32 | 
 33 |     def run(self):
 34 |         logger.debug('loader %d start' % self.thread_id)
 35 |         while True:
 36 |             items = list()
 37 | 
 38 |             for _ in range(self.batch_size):
 39 |                 if self.index >= self.dataset_count:
 40 |                     break
 41 | 
 42 |                 input, label = self.dataset.get_item(self.index)
 43 | 
 44 |                 if input is not None:
 45 |                     items.append((input, label))
 46 | 
 47 |                 self.index += 1
 48 | 
 49 |             if len(items) == 0:
 50 |                 batch = self.create_empty_batch()
 51 |                 self.queue.put(batch)
 52 |                 break
 53 | 
 54 |             random.shuffle(items)
 55 | 
 56 |             batch = self.collate_fn(items)
 57 |             self.queue.put(batch)
 58 | 
 59 |         logger.debug('loader %d stop' % self.thread_id)
 60 | 
 61 | 
 62 | def _collate_fn(batch):
 63 |     """ functions that pad to the maximum sequence length """
 64 | 
 65 |     def seq_length_(p):
 66 |         return len(p[0])
 67 | 
 68 |     def target_length_(p):
 69 |         return len(p[1])
 70 | 
 71 |     seq_lengths = [len(s[0]) for s in batch]
 72 |     target_lengths = [len(s[1]) for s in batch]
 73 | 
 74 |     max_seq_sample = max(batch, key=seq_length_)[0]
 75 |     max_target_sample = max(batch, key=target_length_)[1]
 76 | 
 77 |     max_seq_size = max_seq_sample.size(0)
 78 |     max_target_size = len(max_target_sample)
 79 |     batch_size = len(batch)
 80 | 
 81 |     seqs = torch.zeros(batch_size, max_seq_size).to(torch.long)
 82 |     targets = torch.zeros(batch_size, max_target_size).to(torch.long)
 83 | 
 84 |     from package.definition import PAD_token
 85 |     targets.fill_(PAD_token)
 86 |     seqs.fill_(PAD_token)
 87 | 
 88 |     for idx in range(batch_size):
 89 |         sample = batch[idx]
 90 |         tensor = sample[0]
 91 |         target = sample[1]
 92 | 
 93 |         seqs[idx].narrow(0, 0, len(tensor)).copy_(torch.LongTensor(tensor))
 94 |         targets[idx].narrow(0, 0, len(target)).copy_(torch.LongTensor(target))
 95 | 
 96 |     return seqs, targets, seq_lengths, target_lengths
 97 | 
 98 | 
 99 | class CustomDataset(Dataset):
100 |     def __init__(self, corpus, sos_id, eos_id, batch_size):
101 |         self.corpus = corpus
102 |         self.sos_id = sos_id
103 |         self.eos_id = eos_id
104 |         self.batch_size = batch_size
105 | 
106 |     def get_item(self, index):
107 |         input = get_input(self.corpus[index], self.sos_id)
108 |         label = get_label(self.corpus[index], self.eos_id)
109 | 
110 |         return input, label
111 | 
112 |     def shuffle(self):
113 |         random.shuffle(self.corpus)
114 | 
115 |     def count(self):
116 |         return len(self.corpus)
117 | 
118 | 
119 | def load_label(label_path, encoding='utf-8'):
120 |     char2id = dict()
121 |     id2char = dict()
122 | 
123 |     with open(label_path, 'r', encoding=encoding) as f:
124 |         labels = csv.reader(f, delimiter=',')
125 |         next(labels)
126 | 
127 |         for row in labels:
128 |             char2id[row[1]] = row[0]
129 |             id2char[int(row[0])] = row[1]
130 | 
131 |     return char2id, id2char
132 | 
133 | 
134 | def load_corpus(filepath):
135 |     with open(filepath, 'rb') as f:
136 |         corpus = pickle.load(f)
137 |         corpus = list(corpus['id'])
138 | 
139 |         return corpus
140 | 


--------------------------------------------------------------------------------
/package/definition.py:
--------------------------------------------------------------------------------
 1 | import logging
 2 | import sys
 3 | logger = logging.getLogger('root')
 4 | FORMAT = "[%(asctime)s %(filename)s:%(lineno)s - %(funcName)s()] %(message)s"
 5 | logging.basicConfig(stream=sys.stdout, level=logging.DEBUG, format=FORMAT)
 6 | logger.setLevel(logging.INFO)
 7 | from package.data_loader import load_label
 8 | char2id, id2char = load_label('./data/train_labels.csv', encoding='utf-8')
 9 | SOS_token = int(char2id['<s>'])
10 | EOS_token = int(char2id['</s>'])
11 | PAD_token = int(char2id['_'])
12 | train_dict = {'loss': [], 'cer': []}
13 | valid_dict = {'loss': [], 'cer': []}


--------------------------------------------------------------------------------
/package/evaluator.py:
--------------------------------------------------------------------------------
 1 | import torch
 2 | from package.definition import logger
 3 | 
 4 | 
 5 | def evaluate(model, queue, perplexity, device):
 6 |     logger.info('evaluate() start')
 7 | 
 8 |     total_loss = 0
 9 |     total_num = 0
10 | 
11 |     model.eval()
12 | 
13 |     with torch.no_grad():
14 |         while True:
15 |             loss = perplexity
16 | 
17 |             inputs, targets, input_lengths, target_lengths = queue.get()
18 | 
19 |             if inputs.shape[0] == 0:
20 |                 break
21 | 
22 |             inputs = inputs.to(device)
23 |             targets = targets.to(device)
24 | 
25 |             model.module.flatten_parameters()
26 |             outputs = model(inputs, teacher_forcing_ratio=0.0)
27 | 
28 |             loss.reset()
29 |             for step, step_output in enumerate(outputs):
30 |                 batch_size = targets.size(0)
31 |                 loss.eval_batch(step_output.contiguous().view(batch_size, -1), targets[:, step])
32 | 
33 |             loss = loss.get_loss()
34 | 
35 |             total_loss += loss
36 |             total_num += sum(input_lens)
37 | 
38 |     logger.info('evaluate() completed')
39 | 
40 |     return total_loss / total_num
41 | 


--------------------------------------------------------------------------------
/package/loss.py:
--------------------------------------------------------------------------------
  1 | # Refer to IBM/pytorch-seq2seq
  2 | # https://github.com/IBM/pytorch-seq2seq/blob/master/seq2seq/loss/loss.py
  3 | 
  4 | import torch
  5 | import math
  6 | import torch.nn as nn
  7 | import numpy as np
  8 | 
  9 | 
 10 | class Loss(object):
 11 |     """ Base class for encapsulation of the loss functions.
 12 |     This class defines interfaces that are commonly used with loss functions
 13 |     in training and inferencing.  For information regarding individual loss
 14 |     functions, please refer to http://pytorch.org/docs/master/nn.html#loss-functions
 15 |     Note:
 16 |         Do not use this class directly, use one of the sub classes.
 17 |     Args:
 18 |         name (str): name of the loss function used by logging messages.
 19 |         criterion (torch.nn._Loss): one of PyTorch's loss function.  Refer
 20 |             to http://pytorch.org/docs/master/nn.html#loss-functions for
 21 |             a list of them.
 22 |     Attributes:
 23 |         name (str): name of the loss function used by logging messages.
 24 |         criterion (torch.nn._Loss): one of PyTorch's loss function.  Refer
 25 |             to http://pytorch.org/docs/master/nn.html#loss-functions for
 26 |             a list of them.  Implementation depends on individual
 27 |             sub-classes.
 28 |         acc_loss (int or torcn.nn.Tensor): variable that stores accumulated loss.
 29 |         norm_term (float): normalization term that can be used to calculate
 30 |             the loss of multiple batches.  Implementation depends on individual
 31 |             sub-classes.
 32 |     """
 33 | 
 34 |     def __init__(self, name, criterion):
 35 |         self.name = name
 36 |         self.criterion = criterion
 37 |         if not issubclass(type(self.criterion), nn.modules.loss._Loss):
 38 |             raise ValueError("Criterion has to be a subclass of torch.nn._Loss")
 39 |         # accumulated loss
 40 |         self.acc_loss = 0
 41 |         # normalization term
 42 |         self.norm_term = 0
 43 | 
 44 |     def reset(self):
 45 |         """ Reset the accumulated loss. """
 46 |         self.acc_loss = 0
 47 |         self.norm_term = 0
 48 | 
 49 |     def get_loss(self):
 50 |         """ Get the loss.
 51 |         This method defines how to calculate the averaged loss given the
 52 |         accumulated loss and the normalization term.  Override to define your
 53 |         own logic.
 54 |         Returns:
 55 |             loss (float): value of the loss.
 56 |         """
 57 |         raise NotImplementedError
 58 | 
 59 |     def eval_batch(self, outputs, target):
 60 |         """ Evaluate and accumulate loss given outputs and expected results.
 61 |         This method is called after each batch with the batch outputs and
 62 |         the target (expected) results.  The loss and normalization term are
 63 |         accumulated in this method.  Override it to define your own accumulation
 64 |         method.
 65 |         Args:
 66 |             outputs (torch.Tensor): outputs of a batch.
 67 |             target (torch.Tensor): expected output of a batch.
 68 |         """
 69 |         raise NotImplementedError
 70 | 
 71 |     def cuda(self):
 72 |         self.criterion.cuda()
 73 | 
 74 |     def backward(self):
 75 |         if type(self.acc_loss) is int:
 76 |             raise ValueError("No loss to back propagate.")
 77 |         self.acc_loss.backward()
 78 | 
 79 | 
 80 | class NLLLoss(Loss):
 81 |     """ Batch averaged negative log-likelihood loss.
 82 |     Args:
 83 |         weight (torch.Tensor, optional): refer to http://pytorch.org/docs/master/nn.html#nllloss
 84 |         mask (int, optional): index of masked token, i.e. weight[mask] = 0.
 85 |         size_average (bool, optional): refer to http://pytorch.org/docs/master/nn.html#nllloss
 86 |     """
 87 | 
 88 |     _NAME = "Avg NLLLoss"
 89 | 
 90 |     def __init__(self, weight=None, mask=None, size_average=True):
 91 |         self.mask = mask
 92 |         self.size_average = size_average
 93 |         if mask is not None:
 94 |             if weight is None:
 95 |                 raise ValueError("Must provide weight with a mask.")
 96 |             weight[mask] = 0
 97 | 
 98 |         super(NLLLoss, self).__init__(
 99 |             self._NAME,
100 |             nn.NLLLoss(weight=weight, reduction='sum'))
101 | 
102 |     def get_loss(self):
103 |         if isinstance(self.acc_loss, int):
104 |             return 0
105 |         # total loss for all batches
106 |         loss = self.acc_loss.data.item()
107 |         if self.size_average:
108 |             # average loss per batch
109 |             loss /= self.norm_term
110 |         return loss
111 | 
112 |     def eval_batch(self, outputs, target):
113 |         self.acc_loss += self.criterion(outputs, target)
114 |         self.norm_term += 1
115 | 
116 | 
117 | class Perplexity(NLLLoss):
118 |     """ Language model perplexity loss.
119 |     Perplexity is the token averaged likelihood.  When the averaging options are the
120 |     same, it is the exponential of negative log-likelihood.
121 |     Args:
122 |         weight (torch.Tensor, optional): refer to http://pytorch.org/docs/master/nn.html#nllloss
123 |         mask (int, optional): index of masked token, i.e. weight[mask] = 0.
124 |     """
125 | 
126 |     _NAME = "Perplexity"
127 |     _MAX_EXP = 100
128 | 
129 |     def __init__(self, weight=None, mask=None, device=None):
130 |         super(Perplexity, self).__init__(weight=weight, mask=mask, size_average=False)
131 |         self.device = device
132 | 
133 |     def eval_batch(self, outputs, target):
134 |         self.acc_loss += self.criterion(outputs, target).to(self.device)
135 |         if self.mask is None:
136 |             self.norm_term += np.prod(target.size())
137 |         else:
138 |             self.norm_term += target.data.ne(self.mask).sum()
139 | 
140 |     def get_loss(self):
141 |         nll = super(Perplexity, self).get_loss()
142 |         nll /= self.norm_term.item()
143 |         if nll > Perplexity._MAX_EXP:
144 |             print("WARNING: Loss exceeded maximum value, capping to e^100")
145 |             return math.exp(Perplexity._MAX_EXP)
146 |         return math.exp(nll)
147 | 


--------------------------------------------------------------------------------
/package/trainer.py:
--------------------------------------------------------------------------------
 1 | import time
 2 | import torch
 3 | from package.definition import logger
 4 | 
 5 | train_step_result = {'loss': [], 'cer': []}
 6 | 
 7 | 
 8 | def supervised_train(model, queue, perplexity, optimizer, device, print_every, epoch,
 9 |                      teacher_forcing_ratio, worker_num, total_time_step, train_begin):
10 |     print_loss_total = 0  # Reset every print_every
11 |     epoch_loss_total = 0  # Reset every epoch
12 |     total_num = 0
13 |     time_step = 0
14 | 
15 |     model.train()
16 |     begin = epoch_begin = time.time()
17 | 
18 |     while True:
19 |         loss = perplexity
20 |         inputs, targets, input_lens, target_lens = queue.get()
21 | 
22 |         if inputs.shape[0] == 0:
23 |             # empty feats means closing one loader
24 |             worker_num -= 1
25 |             logger.debug('left train_loader: %d' % worker_num)
26 | 
27 |             if worker_num == 0:
28 |                 break
29 |             else:
30 |                 continue
31 | 
32 |         inputs = inputs.to(device)
33 |         targets = targets.to(device)
34 | 
35 |         model.module.flatten_parameters()
36 |         outputs = model(inputs, teacher_forcing_ratio=teacher_forcing_ratio)
37 | 
38 |         # Get loss
39 |         loss.reset()
40 |         for step, step_output in enumerate(outputs):
41 |             batch_size = targets.size(0)
42 |             loss.eval_batch(step_output.contiguous().view(batch_size, -1), targets[:, step])
43 |         # Backpropagation
44 |         model.zero_grad()
45 |         loss.backward()
46 |         optimizer.step()
47 |         loss = loss.get_loss()
48 | 
49 |         epoch_loss_total += loss
50 |         print_loss_total += loss
51 |         total_num += sum(input_lens)
52 | 
53 |         time_step += 1
54 |         torch.cuda.empty_cache()
55 | 
56 |         if time_step % print_every == 0:
57 |             current = time.time()
58 |             elapsed = current - begin
59 |             epoch_elapsed = (current - epoch_begin) / 60.0
60 |             train_elapsed = (current - train_begin) / 3600.0
61 | 
62 |             logger.info('timestep: {:4d}/{:4d}, perplexity: {:.4f}, elapsed: {:.2f}s {:.2f}m {:.2f}h'.format(
63 |                 time_step,
64 |                 total_time_step,
65 |                 print_loss_total / print_every,
66 |                 elapsed, epoch_elapsed, train_elapsed
67 |             ))
68 |             print_loss_total = 0
69 |             begin = time.time()
70 | 
71 |         if time_step % 50000 == 0:
72 |             torch.save(model, "./data/epoch%s_%s.pt" % (str(epoch), str(time_step)))
73 | 
74 |     logger.info('train() completed')
75 | 
76 |     return epoch_loss_total / total_num
77 | 


--------------------------------------------------------------------------------
/package/utils.py:
--------------------------------------------------------------------------------
 1 | import Levenshtein as Lev
 2 | import torch
 3 | 
 4 | 
 5 | def get_label(script, eos_id):
 6 |     tokens = script.split()
 7 | 
 8 |     label = list()
 9 |     for token in tokens:
10 |         label.append(int(token))
11 |     label.append(int(eos_id))
12 | 
13 |     return label
14 | 
15 | 
16 | def get_input(script, sos_id):
17 |     tokens = script.split()
18 | 
19 |     label = list()
20 |     label.append(int(sos_id))
21 |     for token in tokens:
22 |         label.append(int(token))
23 | 
24 |     return torch.LongTensor(label)
25 | 
26 | 
27 | def label_to_string(labels, id2char, eos_id):
28 |     if len(labels.shape) == 1:
29 |         sentence = str()
30 |         for label in labels:
31 |             if label.item() == eos_id:
32 |                 break
33 |             sentence += id2char[label.item()]
34 |         return sentence
35 | 
36 |     elif len(labels.shape) == 2:
37 |         sentences = list()
38 |         for batch in labels:
39 |             sentence = str()
40 |             for label in batch:
41 |                 if label.item() == eos_id:
42 |                     break
43 |                 sentence += id2char[label.item()]
44 |             sentences.append(sentence)
45 |         return sentences
46 | 
47 |     else:
48 |         raise ValueError("shape Error !!")
49 | 
50 | 
51 | def char_distance(target, y_hat):
52 |     target = target.replace(' ', '')
53 |     y_hat = y_hat.replace(' ', '')
54 | 
55 |     dist = Lev.distance(y_hat, target)
56 |     length = len(target.replace(' ', ''))
57 | 
58 |     return dist, length
59 | 


--------------------------------------------------------------------------------
/preprocess/KoNPron.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "markdown",
  5 |    "metadata": {},
  6 |    "source": [
  7 |     "# KoNPron"
  8 |    ]
  9 |   },
 10 |   {
 11 |    "cell_type": "code",
 12 |    "execution_count": 1,
 13 |    "metadata": {},
 14 |    "outputs": [],
 15 |    "source": [
 16 |     "class KoNPron:\n",
 17 |     "    def __init__(self):\n",
 18 |     "        self.base_digit = ['0','1','2','3','4','5','6','7','8','9']\n",
 19 |     "        self.super_digit = ['⁰','¹','²','³','⁴','⁵','⁶','⁷','⁸','⁹']\n",
 20 |     "        self.small_scale = ['','십','백','천']\n",
 21 |     "        self.large_scale = ['','만 ','억 ','조 ','경 ','해 ']\n",
 22 |     "        self.literal = ['영','일','이','삼','사','오','육','칠','팔','구']\n",
 23 |     "        self.spoken_unit = ['', '하나', '둘', '셋', '넷', '다섯', '여섯', '일곱', '여덟', '아홉']\n",
 24 |     "        self.spoke_tens = ['','열', '스물', '서른', '마흔', '쉰', '예순', '일흔', '여든', '아흔']\n",
 25 |     "        self.sentence = str()\n",
 26 |     "    \n",
 27 |     "    def _detect(self, sentence):\n",
 28 |     "        self.sentence = sentence\n",
 29 |     "        detection_data = list()\n",
 30 |     "        tmp = str()\n",
 31 |     "        \n",
 32 |     "        total_len = len(sentence)\n",
 33 |     "        point_count = 0\n",
 34 |     "        continuous_count = 0\n",
 35 |     "        digit_type = 'vanilla'\n",
 36 |     "        \n",
 37 |     "        detected = False\n",
 38 |     "        zero_started = False\n",
 39 |     "        \n",
 40 |     "        for idx, char in enumerate(sentence):\n",
 41 |     "            if char in self.base_digit:\n",
 42 |     "                if not detected:\n",
 43 |     "                    detected = True\n",
 44 |     "                    if char is '0':\n",
 45 |     "                        zero_started = True\n",
 46 |     "                tmp += char\n",
 47 |     "                continuous_count += 1\n",
 48 |     "                if zero_started and continuous_count>8:\n",
 49 |     "                    digit_type = 'telephone/none'\n",
 50 |     "                if continuous_count>20:\n",
 51 |     "                    digit_type = 'enormous/none'        \n",
 52 |     "            else:\n",
 53 |     "                continous_count = 0\n",
 54 |     "                zero_started = False\n",
 55 |     "                if char == ',':\n",
 56 |     "                    if idx+1 < total_len and idx > 0:\n",
 57 |     "                        if sentence[idx-1] in self.base_digit and sentence[idx+1] in self.base_digit:\n",
 58 |     "                            tmp += char\n",
 59 |     "                elif char == '.':\n",
 60 |     "                    if idx+1 < total_len and idx > 0:\n",
 61 |     "                        if sentence[idx-1] in self.base_digit and sentence[idx+1] in self.base_digit:\n",
 62 |     "                            point_count += 1\n",
 63 |     "                            if point_count == 1:\n",
 64 |     "                                digit_type = 'fraction'\n",
 65 |     "                            if point_count > 1:\n",
 66 |     "                                digit_type = 'version'\n",
 67 |     "                            tmp += char\n",
 68 |     "                elif char == '^':\n",
 69 |     "                    if idx+1 < total_len and idx > 0:\n",
 70 |     "                        if sentence[idx-1] in self.base_digit and sentence[idx+1] in self.base_digit:\n",
 71 |     "                            digit_type += '/square'\n",
 72 |     "                            tmp += char\n",
 73 |     "                        else:\n",
 74 |     "                            if digit_type != 'exception/none':\n",
 75 |     "                                digit_type = 'exception/none'\n",
 76 |     "                                tmp += char\n",
 77 |     "                elif char in self.super_digit:\n",
 78 |     "                    if idx > 0:\n",
 79 |     "                        if sentence[idx-1] in self.base_digit or sentence[idx-1] in self.super_digit:\n",
 80 |     "                            if not 'square' in digit_type:\n",
 81 |     "                                digit_type += '/square'\n",
 82 |     "                            tmp += char\n",
 83 |     "                        else:\n",
 84 |     "                            if digit_type != 'exception/none':\n",
 85 |     "                                digit_type = 'exception/none'\n",
 86 |     "                                tmp += char\n",
 87 |     "                elif char == '·':\n",
 88 |     "                    if idx+1 < total_len and idx > 0:\n",
 89 |     "                        if sentence[idx-1] in self.base_digit and sentence[idx+1] in self.base_digit:\n",
 90 |     "                            digit_type = 'date'\n",
 91 |     "                            tmp += char\n",
 92 |     "                else:\n",
 93 |     "                    if detected:\n",
 94 |     "                        detected = False\n",
 95 |     "                        if '/' not in digit_type:\n",
 96 |     "                            digit_type += '/none'\n",
 97 |     "                        detection_data.append((digit_type, tmp))\n",
 98 |     "                        tmp = str()\n",
 99 |     "                        digit_type = 'vanilla'\n",
100 |     "                        point_count = 0\n",
101 |     "                        continuous_count = 0\n",
102 |     "                    else:\n",
103 |     "                        if tmp:\n",
104 |     "                            detection_data.append((digit_type, tmp))\n",
105 |     "                            tmp = str()\n",
106 |     "                            digit_type = 'vanilla'\n",
107 |     "                            point_count = 0\n",
108 |     "                            continuous_count = 0\n",
109 |     "                        \n",
110 |     "        if detected:\n",
111 |     "            if '/' not in digit_type:\n",
112 |     "                digit_type += '/none'\n",
113 |     "            detection_data.append((digit_type, tmp))\n",
114 |     "            \n",
115 |     "        elif tmp:\n",
116 |     "            detection_data.append((digit_type, tmp))\n",
117 |     "        return detection_data\n",
118 |     "    \n",
119 |     "    def _preprocess(self, detection_data):\n",
120 |     "        preprocessed_data = list()\n",
121 |     "        \n",
122 |     "        for digit_type, target in detection_data:\n",
123 |     "            original = target\n",
124 |     "            target_seq = list()\n",
125 |     "            reading_method = list()\n",
126 |     "            target_len = len(target)\n",
127 |     "\n",
128 |     "            main_type, sub_type = digit_type.split('/')\n",
129 |     "            if main_type == 'exception':\n",
130 |     "                return None\n",
131 |     "                   \n",
132 |     "            if main_type == 'version':\n",
133 |     "                splited = target.split('.')\n",
134 |     "                for count, frag in enumerate(splited):\n",
135 |     "                    target_seq.append(frag)\n",
136 |     "                    reading_method.append('individual')\n",
137 |     "                    if count < len(splited) - 1:\n",
138 |     "                        target_seq.append('.')\n",
139 |     "                        reading_method.append('point')\n",
140 |     "                preprocessed_data.append((reading_method, target_seq, original))\n",
141 |     "                \n",
142 |     "            if main_type == 'date':\n",
143 |     "                target = target.split('·')\n",
144 |     "                for frag in target:\n",
145 |     "                    target_seq.append(frag)\n",
146 |     "                    reading_method.append('individual')\n",
147 |     "                preprocessed_data.append((reading_method, target_seq, original))\n",
148 |     "                \n",
149 |     "            if main_type == 'telephone' or main_type == 'enormous':\n",
150 |     "                target = target.replace('0', '_')\n",
151 |     "                target_seq.append(target)\n",
152 |     "                reading_method.append('individual')\n",
153 |     "                preprocessed_data.append((reading_method, target_seq, original))\n",
154 |     "                \n",
155 |     "            if sub_type != 'none':\n",
156 |     "                if main_type == 'vanilla':\n",
157 |     "                    target = [target.replace(',','')]\n",
158 |     "                if main_type == 'fraction':\n",
159 |     "                    target = target.split('.')\n",
160 |     "                    \n",
161 |     "                if sub_type == 'square':\n",
162 |     "                    if '^' not in target[-1]:\n",
163 |     "                        super_part = str()\n",
164 |     "                        for idx, digit in enumerate(target[-1]):\n",
165 |     "                            if digit not in self.base_digit:\n",
166 |     "                                super_idx = idx\n",
167 |     "                                break\n",
168 |     "                        super_num = list(target[-1][idx:])\n",
169 |     "                        for idx in range(len(super_num)):\n",
170 |     "                            super_num[idx] = str(self.super_digit.index(super_num[idx]))\n",
171 |     "                        super_num = ''.join(super_num)\n",
172 |     "                        super_part += target[-1][:super_idx]+'^'+super_num\n",
173 |     "                        if main_type == 'vanilla':\n",
174 |     "                            target = [super_part]\n",
175 |     "                        else:\n",
176 |     "                            target = target[:1]+[super_part]\n",
177 |     "                    if main_type == 'fraction':\n",
178 |     "                        target_seq.append(target[0])\n",
179 |     "                        reading_method.append('literal')\n",
180 |     "                        target_seq.append('.')\n",
181 |     "                        reading_method.append('point')\n",
182 |     "\n",
183 |     "                    tmp_len = len(target_seq)\n",
184 |     "                    tmp = target[-1].split('^')\n",
185 |     "                    for seq in tmp:\n",
186 |     "                        target_seq.append(seq)\n",
187 |     "                        if 'point' in reading_method:\n",
188 |     "                            if 'of' in reading_method:\n",
189 |     "                                reading_method.append('literal')\n",
190 |     "                            else:\n",
191 |     "                                reading_method.append('individual')\n",
192 |     "                        else:\n",
193 |     "                            reading_method.append('literal')\n",
194 |     "                        if len(target_seq) == tmp_len+1:\n",
195 |     "                            target_seq.append(\"^\")\n",
196 |     "                            reading_method.append(\"of\")\n",
197 |     "                        elif len(target_seq) == tmp_len+3:\n",
198 |     "                            target_seq.append(\"^\")\n",
199 |     "                            reading_method.append(\"super\")                       \n",
200 |     "                    preprocessed_data.append((reading_method, target_seq, original))\n",
201 |     "            \n",
202 |     "            else:\n",
203 |     "                if main_type == 'vanilla':\n",
204 |     "                    target = target.replace(',','')\n",
205 |     "                    target_seq.append(target)\n",
206 |     "                    reading_method.append('literal')\n",
207 |     "                    preprocessed_data.append((reading_method, target_seq, original))\n",
208 |     "                    \n",
209 |     "                if main_type == 'fraction':\n",
210 |     "                    target = target.split('.')\n",
211 |     "                    for frag in target:\n",
212 |     "                        target_seq.append(frag)\n",
213 |     "                        if 'point' in reading_method:\n",
214 |     "                            reading_method.append('individual')\n",
215 |     "                        else:\n",
216 |     "                            reading_method.append('literal')\n",
217 |     "                        if len(target_seq) == 1:\n",
218 |     "                            target_seq.append(\".\")\n",
219 |     "                            reading_method.append('point')\n",
220 |     "                            \n",
221 |     "                    preprocessed_data.append((reading_method, target_seq, original))\n",
222 |     "                \n",
223 |     "        return preprocessed_data\n",
224 |     "            \n",
225 |     "    \n",
226 |     "    def _read(self, preprocessed_data, mode = 'informal'):\n",
227 |     "        def __literal_read(self, frag, mode = 'informal'):\n",
228 |     "            korean = str()\n",
229 |     "            tmp = str()\n",
230 |     "            length = len(frag)\n",
231 |     "            for idx, digit in enumerate(frag):\n",
232 |     "                digit = int(digit)\n",
233 |     "                inversed_idx = length-idx-1\n",
234 |     "                if mode == 'formal':\n",
235 |     "                    if inversed_idx%4:\n",
236 |     "                        if digit:\n",
237 |     "                            tmp += self.literal[digit]\n",
238 |     "                            tmp += self.small_scale[inversed_idx%4]\n",
239 |     "                    else:\n",
240 |     "                        if digit or length == 1:\n",
241 |     "                            tmp += self.literal[digit]\n",
242 |     "                        if tmp:\n",
243 |     "                            tmp += self.large_scale[inversed_idx//4]\n",
244 |     "                        korean += tmp\n",
245 |     "                        tmp = str()\n",
246 |     "                        \n",
247 |     "                elif mode == 'informal':\n",
248 |     "                    if inversed_idx%4:\n",
249 |     "                        if digit>1:\n",
250 |     "                            tmp += self.literal[digit]\n",
251 |     "                        if digit:\n",
252 |     "                            tmp += self.small_scale[inversed_idx%4]\n",
253 |     "                    else:\n",
254 |     "                        if digit or length == 1:\n",
255 |     "                            tmp += self.literal[digit]\n",
256 |     "                        if tmp:\n",
257 |     "                            tmp += self.large_scale[inversed_idx//4]\n",
258 |     "                        if length == 5 and digit == 1 and inversed_idx == 4:\n",
259 |     "                            tmp = tmp[1:]\n",
260 |     "                        korean += tmp\n",
261 |     "                        tmp = str()\n",
262 |     "            korean += tmp\n",
263 |     "            return korean\n",
264 |     "        \n",
265 |     "        def __individual_read(self, frag):\n",
266 |     "            korean = str()\n",
267 |     "            for digit in frag:\n",
268 |     "                if digit == '_':\n",
269 |     "                    korean += '공'\n",
270 |     "                else:\n",
271 |     "                    korean += self.literal[int(digit)]\n",
272 |     "            return korean\n",
273 |     "        \n",
274 |     "        \n",
275 |     "        result = self.sentence\n",
276 |     "        if preprocessed_data is None:\n",
277 |     "            return None\n",
278 |     "                            \n",
279 |     "        for each in preprocessed_data:\n",
280 |     "            reading_method, target_seq, original = each\n",
281 |     "            readed = str()\n",
282 |     "            for idx, frag in enumerate(target_seq):\n",
283 |     "                if reading_method[idx] == 'literal':\n",
284 |     "                    readed += __literal_read(self, frag)\n",
285 |     "                if reading_method[idx] == 'individual':\n",
286 |     "                    readed += __individual_read(self, frag)\n",
287 |     "                if reading_method[idx] == 'point':\n",
288 |     "                    readed += ' 점 '\n",
289 |     "                if reading_method[idx] == 'of':\n",
290 |     "                    readed += '의 '\n",
291 |     "                if reading_method[idx] == 'super':\n",
292 |     "                    readed += ' 승'\n",
293 |     "            result = result.replace(original, readed, 1)\n",
294 |     "        return result\n",
295 |     "    \n",
296 |     "    def convert(self, sentence):\n",
297 |     "        return self._read(self._preprocess(self._detect(sentence)))"
298 |    ]
299 |   },
300 |   {
301 |    "cell_type": "markdown",
302 |    "metadata": {},
303 |    "source": [
304 |     "## 중간 출력 값 확인"
305 |    ]
306 |   },
307 |   {
308 |    "cell_type": "code",
309 |    "execution_count": 12,
310 |    "metadata": {},
311 |    "outputs": [
312 |     {
313 |      "name": "stdout",
314 |      "output_type": "stream",
315 |      "text": [
316 |       "Original :  13,000, 3.14, 1.4^2, 8·15, 1.1.7\n",
317 |       "<<detect>>\n",
318 |       "[('vanilla/none', '13,000'), ('fraction/none', '3.14'), ('fraction/square', '1.4^2'), ('date/none', '8·15'), ('version/none', '1.1.7')]\n",
319 |       "<<preprocessed>>\n",
320 |       "[(['literal'], ['13000'], '13,000'), (['literal', 'point', 'individual'], ['3', '.', '14'], '3.14'), (['literal', 'point', 'individual', 'of', 'literal', 'super'], ['1', '.', '4', '^', '2', '^'], '1.4^2'), (['individual', 'individual'], ['8', '15'], '8·15'), (['individual', 'point', 'individual', 'point', 'individual'], ['1', '.', '1', '.', '7'], '1.1.7')]\n",
321 |       "<<read>>\n",
322 |       "만 삼천, 삼 점 일사, 일 점 사의 이 승, 팔일오, 일 점 일 점 칠\n"
323 |      ]
324 |     }
325 |    ],
326 |    "source": [
327 |     "test = KoNPron()\n",
328 |     "sentence = \"01066883225, 420076049241990172380904740935\"\n",
329 |     "sentence = \"13,000, 3.14, 1.4^2, 8·15, 1.1.7\"\n",
330 |     "print(\"Original : \",sentence)\n",
331 |     "detect = test._detect(sentence)\n",
332 |     "print(\"<<detect>>\")\n",
333 |     "print(detect)\n",
334 |     "preprocessed = test._preprocess(detect)\n",
335 |     "print(\"<<preprocessed>>\")\n",
336 |     "print(preprocessed)\n",
337 |     "print(\"<<read>>\")\n",
338 |     "print(test._read(preprocessed))"
339 |    ]
340 |   },
341 |   {
342 |    "cell_type": "markdown",
343 |    "metadata": {},
344 |    "source": [
345 |     "## 예외상황에 대하여는 중간출력 및 최종출력을 None으로 설정"
346 |    ]
347 |   },
348 |   {
349 |    "cell_type": "code",
350 |    "execution_count": 3,
351 |    "metadata": {},
352 |    "outputs": [
353 |     {
354 |      "name": "stdout",
355 |      "output_type": "stream",
356 |      "text": [
357 |       "[('exception/none', '⁴')]\n",
358 |       "None\n",
359 |       "None\n"
360 |      ]
361 |     }
362 |    ],
363 |    "source": [
364 |     "exception = \"아니⁴ 이런 일이?\"\n",
365 |     "detect = test._detect(exception)\n",
366 |     "print(detect)\n",
367 |     "preprocessed = test._preprocess(detect)\n",
368 |     "print(preprocessed)\n",
369 |     "result = test._read(preprocessed)\n",
370 |     "print(result)"
371 |    ]
372 |   },
373 |   {
374 |    "cell_type": "markdown",
375 |    "metadata": {},
376 |    "source": [
377 |     "## 예외 문장과 정상 문장을 구분하여 출력"
378 |    ]
379 |   },
380 |   {
381 |    "cell_type": "code",
382 |    "execution_count": 6,
383 |    "metadata": {},
384 |    "outputs": [
385 |     {
386 |      "name": "stdout",
387 |      "output_type": "stream",
388 |      "text": [
389 |       "팔일오 행사를 맞아 광장에 사만 구천여 명의 시민들이 모였다. 이 중 오십육 점 이이퍼센트는 태극기를 들고 있었다.\n",
390 |       "버전 일 점 영 점 칠, 삼천만큼 사랑해, 십육 점 칠의 삼 승과 만 오천삼백삼십이 점 칠팔삼의 백십 승\n",
391 |       "제 핸드폰 번호는 공일공일이삼사오육칠팔, 집 전화번호는 공이일이삼사오육칠이고 송장번호는 사이공공칠육공사구이사일구구공일칠이삼팔공구공사칠사공구삼오\n"
392 |      ]
393 |     }
394 |    ],
395 |    "source": [
396 |     "sentence1 = \"8·15 행사를 맞아 광장에 4,9000여 명의 시민들이 모였다. 이 중 56.22퍼센트는 태극기를 들고 있었다.\"\n",
397 |     "sentence2 = \"아니 벌써³ 4월이라니\"\n",
398 |     "exception1 = \"그러게요^^\"\n",
399 |     "exception2 = \"버전 1.0.7, 3000만큼 사랑해, 16.7³과 15332.783^110\"\n",
400 |     "sentence3 = \"제 핸드폰 번호는 01012345678, 집 전화번호는 021234567이고 송장번호는 420076049241990172380904740935\"\n",
401 |     "\n",
402 |     "data = [sentence1, exception1, exception2, sentence2, sentence3]\n",
403 |     "\n",
404 |     "for sentence in data:\n",
405 |     "    ret = test.convert(sentence)\n",
406 |     "    if ret is not None:\n",
407 |     "        print(ret)"
408 |    ]
409 |   },
410 |   {
411 |    "cell_type": "code",
412 |    "execution_count": null,
413 |    "metadata": {},
414 |    "outputs": [],
415 |    "source": []
416 |   }
417 |  ],
418 |  "metadata": {
419 |   "kernelspec": {
420 |    "display_name": "Python 3",
421 |    "language": "python",
422 |    "name": "python3"
423 |   },
424 |   "language_info": {
425 |    "codemirror_mode": {
426 |     "name": "ipython",
427 |     "version": 3
428 |    },
429 |    "file_extension": ".py",
430 |    "mimetype": "text/x-python",
431 |    "name": "python",
432 |    "nbconvert_exporter": "python",
433 |    "pygments_lexer": "ipython3",
434 |    "version": "3.7.3"
435 |   }
436 |  },
437 |  "nbformat": 4,
438 |  "nbformat_minor": 2
439 | }
440 | 


--------------------------------------------------------------------------------
/preprocess/Preprocess.ipynb:
--------------------------------------------------------------------------------
   1 | {
   2 |  "cells": [
   3 |   {
   4 |    "cell_type": "markdown",
   5 |    "metadata": {},
   6 |    "source": [
   7 |     "# Preprocess for Language Model"
   8 |    ]
   9 |   },
  10 |   {
  11 |    "cell_type": "markdown",
  12 |    "metadata": {},
  13 |    "source": [
  14 |     "수집한 Corpus를 통합하여 일련의 규칙으로 전처리"
  15 |    ]
  16 |   },
  17 |   {
  18 |    "cell_type": "markdown",
  19 |    "metadata": {},
  20 |    "source": [
  21 |     "* 필요한 라이브러리 import"
  22 |    ]
  23 |   },
  24 |   {
  25 |    "cell_type": "code",
  26 |    "execution_count": 1,
  27 |    "metadata": {},
  28 |    "outputs": [],
  29 |    "source": [
  30 |     "import pandas as pd\n",
  31 |     "import numpy as np\n",
  32 |     "import random\n",
  33 |     "import pickle"
  34 |    ]
  35 |   },
  36 |   {
  37 |    "cell_type": "markdown",
  38 |    "metadata": {},
  39 |    "source": [
  40 |     "* 수집한 corpus 통합"
  41 |    ]
  42 |   },
  43 |   {
  44 |    "cell_type": "code",
  45 |    "execution_count": 2,
  46 |    "metadata": {},
  47 |    "outputs": [
  48 |     {
  49 |      "ename": "FileNotFoundError",
  50 |      "evalue": "[Errno 2] No such file or directory: '../data/OpenSubtitles.csv'",
  51 |      "output_type": "error",
  52 |      "traceback": [
  53 |       "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
  54 |       "\u001b[1;31mFileNotFoundError\u001b[0m                         Traceback (most recent call last)",
  55 |       "\u001b[1;32m<ipython-input-2-183ad78126ec>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mkor_corpus\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mread_csv\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'../data/OpenSubtitles.csv'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mencoding\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'utf-8'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[0msub_corpus\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mread_csv\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'../data/new_kor.csv'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mencoding\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'cp949'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[0mkor_corpus\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mconcat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkor_corpus\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msub_corpus\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0msub_corpus\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mread_csv\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'../data/tatoeba.csv'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mencoding\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'utf-8'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
  56 |       "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\pandas\\io\\parsers.py\u001b[0m in \u001b[0;36mparser_f\u001b[1;34m(filepath_or_buffer, sep, delimiter, header, names, index_col, usecols, squeeze, prefix, mangle_dupe_cols, dtype, engine, converters, true_values, false_values, skipinitialspace, skiprows, skipfooter, nrows, na_values, keep_default_na, na_filter, verbose, skip_blank_lines, parse_dates, infer_datetime_format, keep_date_col, date_parser, dayfirst, cache_dates, iterator, chunksize, compression, thousands, decimal, lineterminator, quotechar, quoting, doublequote, escapechar, comment, encoding, dialect, error_bad_lines, warn_bad_lines, delim_whitespace, low_memory, memory_map, float_precision)\u001b[0m\n\u001b[0;32m    674\u001b[0m         )\n\u001b[0;32m    675\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 676\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0m_read\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfilepath_or_buffer\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mkwds\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    677\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    678\u001b[0m     \u001b[0mparser_f\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m__name__\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mname\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
  57 |       "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\pandas\\io\\parsers.py\u001b[0m in \u001b[0;36m_read\u001b[1;34m(filepath_or_buffer, kwds)\u001b[0m\n\u001b[0;32m    446\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    447\u001b[0m     \u001b[1;31m# Create the parser.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 448\u001b[1;33m     \u001b[0mparser\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mTextFileReader\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfp_or_buf\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    449\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    450\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0mchunksize\u001b[0m \u001b[1;32mor\u001b[0m \u001b[0miterator\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
  58 |       "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\pandas\\io\\parsers.py\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, f, engine, **kwds)\u001b[0m\n\u001b[0;32m    878\u001b[0m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0moptions\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m\"has_index_names\"\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mkwds\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m\"has_index_names\"\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    879\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 880\u001b[1;33m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_make_engine\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mengine\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    881\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    882\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mclose\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
  59 |       "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\pandas\\io\\parsers.py\u001b[0m in \u001b[0;36m_make_engine\u001b[1;34m(self, engine)\u001b[0m\n\u001b[0;32m   1112\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_make_engine\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mengine\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m\"c\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1113\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mengine\u001b[0m \u001b[1;33m==\u001b[0m \u001b[1;34m\"c\"\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1114\u001b[1;33m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mCParserWrapper\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mf\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0moptions\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1115\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1116\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mengine\u001b[0m \u001b[1;33m==\u001b[0m \u001b[1;34m\"python\"\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
  60 |       "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\pandas\\io\\parsers.py\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, src, **kwds)\u001b[0m\n\u001b[0;32m   1872\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mkwds\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"compression\"\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0mencoding\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1873\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msrc\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mstr\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1874\u001b[1;33m                 \u001b[0msrc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mopen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msrc\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m\"rb\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1875\u001b[0m                 \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mhandles\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msrc\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1876\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
  61 |       "\u001b[1;31mFileNotFoundError\u001b[0m: [Errno 2] No such file or directory: '../data/OpenSubtitles.csv'"
  62 |      ]
  63 |     }
  64 |    ],
  65 |    "source": [
  66 |     "kor_corpus = pd.read_csv('../data/OpenSubtitles.csv', encoding='utf-8')\n",
  67 |     "sub_corpus = pd.read_csv('../data/new_kor.csv', encoding='cp949')\n",
  68 |     "kor_corpus = pd.concat([kor_corpus, sub_corpus])\n",
  69 |     "\n",
  70 |     "sub_corpus = pd.read_csv('../data/tatoeba.csv', encoding='utf-8')\n",
  71 |     "kor_corpus = pd.concat([kor_corpus, sub_corpus])\n",
  72 |     "\n",
  73 |     "sub_corpus = pd.read_csv('../data/네이버 해커톤 데이터셋.csv', encoding='utf-8', engine='python')\n",
  74 |     "kor_corpus = pd.concat([kor_corpus, sub_corpus])\n",
  75 |     "\n",
  76 |     "for i in range(10):\n",
  77 |     "    sub_corpus = pd.read_csv('../data/AI Hub 한-영 데이터셋[%s].csv' % i, encoding='utf-8', engine='python')\n",
  78 |     "    kor_corpus = pd.concat([kor_corpus, sub_corpus])"
  79 |    ]
  80 |   },
  81 |   {
  82 |    "cell_type": "markdown",
  83 |    "metadata": {},
  84 |    "source": [
  85 |     "* corpus 크기 확인"
  86 |    ]
  87 |   },
  88 |   {
  89 |    "cell_type": "code",
  90 |    "execution_count": 3,
  91 |    "metadata": {},
  92 |    "outputs": [
  93 |     {
  94 |      "data": {
  95 |       "text/plain": [
  96 |        "3048264"
  97 |       ]
  98 |      },
  99 |      "execution_count": 3,
 100 |      "metadata": {},
 101 |      "output_type": "execute_result"
 102 |     }
 103 |    ],
 104 |    "source": [
 105 |     "len(kor_corpus)"
 106 |    ]
 107 |   },
 108 |   {
 109 |    "cell_type": "markdown",
 110 |    "metadata": {},
 111 |    "source": [
 112 |     "* 코퍼스 내용 확인"
 113 |    ]
 114 |   },
 115 |   {
 116 |    "cell_type": "code",
 117 |    "execution_count": 4,
 118 |    "metadata": {},
 119 |    "outputs": [
 120 |     {
 121 |      "data": {
 122 |       "text/html": [
 123 |        "<div>\n",
 124 |        "<style scoped>\n",
 125 |        "    .dataframe tbody tr th:only-of-type {\n",
 126 |        "        vertical-align: middle;\n",
 127 |        "    }\n",
 128 |        "\n",
 129 |        "    .dataframe tbody tr th {\n",
 130 |        "        vertical-align: top;\n",
 131 |        "    }\n",
 132 |        "\n",
 133 |        "    .dataframe thead th {\n",
 134 |        "        text-align: right;\n",
 135 |        "    }\n",
 136 |        "</style>\n",
 137 |        "<table border=\"1\" class=\"dataframe\">\n",
 138 |        "  <thead>\n",
 139 |        "    <tr style=\"text-align: right;\">\n",
 140 |        "      <th></th>\n",
 141 |        "      <th>ko</th>\n",
 142 |        "    </tr>\n",
 143 |        "  </thead>\n",
 144 |        "  <tbody>\n",
 145 |        "    <tr>\n",
 146 |        "      <th>0</th>\n",
 147 |        "      <td>폭설이 내리고 우박, 진눈깨비가 퍼부어도 눈보라가 몰아쳐도 강풍이 불고 비바람이 휘...</td>\n",
 148 |        "    </tr>\n",
 149 |        "    <tr>\n",
 150 |        "      <th>1</th>\n",
 151 |        "      <td>우리의 한결같은 심부름꾼 황새 아저씨 가는 길을 그 누가 막으랴!</td>\n",
 152 |        "    </tr>\n",
 153 |        "    <tr>\n",
 154 |        "      <th>2</th>\n",
 155 |        "      <td>황새 아저씨를 기다리세요</td>\n",
 156 |        "    </tr>\n",
 157 |        "    <tr>\n",
 158 |        "      <th>3</th>\n",
 159 |        "      <td>찾아와 선물을 주실 거예요</td>\n",
 160 |        "    </tr>\n",
 161 |        "    <tr>\n",
 162 |        "      <th>4</th>\n",
 163 |        "      <td>가난하든 부자이든 상관이 없답니다</td>\n",
 164 |        "    </tr>\n",
 165 |        "  </tbody>\n",
 166 |        "</table>\n",
 167 |        "</div>"
 168 |       ],
 169 |       "text/plain": [
 170 |        "                                                  ko\n",
 171 |        "0  폭설이 내리고 우박, 진눈깨비가 퍼부어도 눈보라가 몰아쳐도 강풍이 불고 비바람이 휘...\n",
 172 |        "1               우리의 한결같은 심부름꾼 황새 아저씨 가는 길을 그 누가 막으랴!\n",
 173 |        "2                                      황새 아저씨를 기다리세요\n",
 174 |        "3                                     찾아와 선물을 주실 거예요\n",
 175 |        "4                                 가난하든 부자이든 상관이 없답니다"
 176 |       ]
 177 |      },
 178 |      "execution_count": 4,
 179 |      "metadata": {},
 180 |      "output_type": "execute_result"
 181 |     }
 182 |    ],
 183 |    "source": [
 184 |     "kor_corpus.head()"
 185 |    ]
 186 |   },
 187 |   {
 188 |    "cell_type": "markdown",
 189 |    "metadata": {},
 190 |    "source": [
 191 |     "* list로 변환 및 shuffle (기존에는 그냥 concatenate 했기 때문에 섞어줌)"
 192 |    ]
 193 |   },
 194 |   {
 195 |    "cell_type": "code",
 196 |    "execution_count": 5,
 197 |    "metadata": {},
 198 |    "outputs": [
 199 |     {
 200 |      "data": {
 201 |       "text/plain": [
 202 |        "['어디 보자...',\n",
 203 |        " '7대 왕국 종족 중에 오크는 처음 듣는다',\n",
 204 |        " '이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.',\n",
 205 |        " '라니스터 가의',\n",
 206 |        " '별 희한한 생각이 다 떠오르곤 하죠',\n",
 207 |        " '아!',\n",
 208 |        " '회의 전에 마실 것 좀 드릴까요? 커피와 차가 있습니다.',\n",
 209 |        " '나는 그런 일에는 전혀 흥미가 없어요.',\n",
 210 |        " '일자리가 있고 제대로 된 회사가 있다면 늘 사람들이 붐비고 활력이 넘치는 도시로 자리매김할 수 있다.',\n",
 211 |        " '카카오가 많이 함유된 다크 초콜릿이 더 좋다는 것을 잊지 말자.',\n",
 212 |        " '비오 13세는 누구인가?',\n",
 213 |        " '네, 어..',\n",
 214 |        " '치밀한 사람이긴 하지만...',\n",
 215 |        " '계속 절 따라오셨죠',\n",
 216 |        " '세계 4위 특허출원 강국이지만 지식재산 심사 품질과 보호 수준이 낮아 가치를 제대로 인정받지 못하고 있다.',\n",
 217 |        " '나는 전부터 이 학교를 잘 알고 있었어.',\n",
 218 |        " '-뭔가 꿍꿍이를 꾸미는게 틀림없어',\n",
 219 |        " '어쩔 수 없이 과학 수업에 늦었습니다.',\n",
 220 |        " '가르시아 에르난데스, 토니 그란데, 하비에르 미냐노 등 축구대표팀의 외국인 코치들이 8일 오후 (현지시간) 오스트리아 레오강의 대표팀 숙소에서 인터뷰에 응하고있다.',\n",
 221 |        " '검찰은 “(증인신문 내용이) 다 녹음 됐으니까 (이 전 대통령이 한 말에 대해) 따지려면 따져볼 수 있다”고 말하기도 했다.']"
 222 |       ]
 223 |      },
 224 |      "execution_count": 5,
 225 |      "metadata": {},
 226 |      "output_type": "execute_result"
 227 |     }
 228 |    ],
 229 |    "source": [
 230 |     "kor_corpus = list(kor_corpus['ko'])\n",
 231 |     "random.shuffle(kor_corpus)\n",
 232 |     "kor_corpus[:20]"
 233 |    ]
 234 |   },
 235 |   {
 236 |    "cell_type": "markdown",
 237 |    "metadata": {},
 238 |    "source": [
 239 |     "* ( ) 와 같이 괄호 처리된 내용들을 필터링한다. (Speech-To-Text에 사용할 언어모델이기 때문에 괄호 및 괄호 내용은 필요없음)"
 240 |    ]
 241 |   },
 242 |   {
 243 |    "cell_type": "code",
 244 |    "execution_count": 6,
 245 |    "metadata": {},
 246 |    "outputs": [],
 247 |    "source": [
 248 |     "new_corpus = list()\n",
 249 |     "flag = False\n",
 250 |     "\n",
 251 |     "for sentence in kor_corpus:\n",
 252 |     "    new_sentence = str()\n",
 253 |     "    for ch in sentence:\n",
 254 |     "        if ch == '(':\n",
 255 |     "            flag = True\n",
 256 |     "            continue\n",
 257 |     "        elif ch == ')':\n",
 258 |     "            flag = False\n",
 259 |     "            continue\n",
 260 |     "        elif flag == False:\n",
 261 |     "            new_sentence += ch\n",
 262 |     "        else:\n",
 263 |     "            continue\n",
 264 |     "            \n",
 265 |     "    new_corpus.append(new_sentence)"
 266 |    ]
 267 |   },
 268 |   {
 269 |    "cell_type": "markdown",
 270 |    "metadata": {},
 271 |    "source": [
 272 |     "* ()가 잘 필터링 됐는지 확인"
 273 |    ]
 274 |   },
 275 |   {
 276 |    "cell_type": "code",
 277 |    "execution_count": 7,
 278 |    "metadata": {},
 279 |    "outputs": [
 280 |     {
 281 |      "data": {
 282 |       "text/plain": [
 283 |        "['어디 보자...',\n",
 284 |        " '7대 왕국 종족 중에 오크는 처음 듣는다',\n",
 285 |        " '이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.',\n",
 286 |        " '라니스터 가의',\n",
 287 |        " '별 희한한 생각이 다 떠오르곤 하죠',\n",
 288 |        " '아!',\n",
 289 |        " '회의 전에 마실 것 좀 드릴까요? 커피와 차가 있습니다.',\n",
 290 |        " '나는 그런 일에는 전혀 흥미가 없어요.',\n",
 291 |        " '일자리가 있고 제대로 된 회사가 있다면 늘 사람들이 붐비고 활력이 넘치는 도시로 자리매김할 수 있다.',\n",
 292 |        " '카카오가 많이 함유된 다크 초콜릿이 더 좋다는 것을 잊지 말자.',\n",
 293 |        " '비오 13세는 누구인가?',\n",
 294 |        " '네, 어..',\n",
 295 |        " '치밀한 사람이긴 하지만...',\n",
 296 |        " '계속 절 따라오셨죠',\n",
 297 |        " '세계 4위 특허출원 강국이지만 지식재산 심사 품질과 보호 수준이 낮아 가치를 제대로 인정받지 못하고 있다.',\n",
 298 |        " '나는 전부터 이 학교를 잘 알고 있었어.',\n",
 299 |        " '-뭔가 꿍꿍이를 꾸미는게 틀림없어',\n",
 300 |        " '어쩔 수 없이 과학 수업에 늦었습니다.',\n",
 301 |        " '가르시아 에르난데스, 토니 그란데, 하비에르 미냐노 등 축구대표팀의 외국인 코치들이 8일 오후  오스트리아 레오강의 대표팀 숙소에서 인터뷰에 응하고있다.',\n",
 302 |        " '검찰은 “ 다 녹음 됐으니까  따지려면 따져볼 수 있다”고 말하기도 했다.']"
 303 |       ]
 304 |      },
 305 |      "execution_count": 7,
 306 |      "metadata": {},
 307 |      "output_type": "execute_result"
 308 |     }
 309 |    ],
 310 |    "source": [
 311 |     "kor_corpus = new_corpus\n",
 312 |     "kor_corpus[:20]"
 313 |    ]
 314 |   },
 315 |   {
 316 |    "cell_type": "markdown",
 317 |    "metadata": {},
 318 |    "source": [
 319 |     "* 현재 corpus에 어떤 특수문자들이 있는지 확인"
 320 |    ]
 321 |   },
 322 |   {
 323 |    "cell_type": "code",
 324 |    "execution_count": 8,
 325 |    "metadata": {},
 326 |    "outputs": [],
 327 |    "source": [
 328 |     "special_ch = list()\n",
 329 |     "\n",
 330 |     "for sentence in kor_corpus:\n",
 331 |     "    for ch in sentence:\n",
 332 |     "        if ch.isdigit() == False and ch.isalpha() == False and ch not in special_ch:\n",
 333 |     "            special_ch.append(ch)"
 334 |    ]
 335 |   },
 336 |   {
 337 |    "cell_type": "markdown",
 338 |    "metadata": {},
 339 |    "source": [
 340 |     "* corpus에 등장한 특수문자 개수 확인"
 341 |    ]
 342 |   },
 343 |   {
 344 |    "cell_type": "code",
 345 |    "execution_count": 9,
 346 |    "metadata": {},
 347 |    "outputs": [
 348 |     {
 349 |      "data": {
 350 |       "text/plain": [
 351 |        "209"
 352 |       ]
 353 |      },
 354 |      "execution_count": 9,
 355 |      "metadata": {},
 356 |      "output_type": "execute_result"
 357 |     }
 358 |    ],
 359 |    "source": [
 360 |     "len(special_ch)"
 361 |    ]
 362 |   },
 363 |   {
 364 |    "cell_type": "markdown",
 365 |    "metadata": {},
 366 |    "source": [
 367 |     "* 등장한 특수문자를 csv 파일로 저장하여 적절한 발음전사로 매핑"
 368 |    ]
 369 |   },
 370 |   {
 371 |    "cell_type": "code",
 372 |    "execution_count": 10,
 373 |    "metadata": {},
 374 |    "outputs": [
 375 |     {
 376 |      "data": {
 377 |       "text/html": [
 378 |        "<div>\n",
 379 |        "<style scoped>\n",
 380 |        "    .dataframe tbody tr th:only-of-type {\n",
 381 |        "        vertical-align: middle;\n",
 382 |        "    }\n",
 383 |        "\n",
 384 |        "    .dataframe tbody tr th {\n",
 385 |        "        vertical-align: top;\n",
 386 |        "    }\n",
 387 |        "\n",
 388 |        "    .dataframe thead th {\n",
 389 |        "        text-align: right;\n",
 390 |        "    }\n",
 391 |        "</style>\n",
 392 |        "<table border=\"1\" class=\"dataframe\">\n",
 393 |        "  <thead>\n",
 394 |        "    <tr style=\"text-align: right;\">\n",
 395 |        "      <th></th>\n",
 396 |        "      <th>special</th>\n",
 397 |        "      <th>replace</th>\n",
 398 |        "    </tr>\n",
 399 |        "  </thead>\n",
 400 |        "  <tbody>\n",
 401 |        "    <tr>\n",
 402 |        "      <th>0</th>\n",
 403 |        "      <td></td>\n",
 404 |        "      <td></td>\n",
 405 |        "    </tr>\n",
 406 |        "    <tr>\n",
 407 |        "      <th>1</th>\n",
 408 |        "      <td>.</td>\n",
 409 |        "      <td>.</td>\n",
 410 |        "    </tr>\n",
 411 |        "    <tr>\n",
 412 |        "      <th>2</th>\n",
 413 |        "      <td>“</td>\n",
 414 |        "      <td></td>\n",
 415 |        "    </tr>\n",
 416 |        "    <tr>\n",
 417 |        "      <th>3</th>\n",
 418 |        "      <td>\"</td>\n",
 419 |        "      <td></td>\n",
 420 |        "    </tr>\n",
 421 |        "    <tr>\n",
 422 |        "      <th>4</th>\n",
 423 |        "      <td>-</td>\n",
 424 |        "      <td></td>\n",
 425 |        "    </tr>\n",
 426 |        "  </tbody>\n",
 427 |        "</table>\n",
 428 |        "</div>"
 429 |       ],
 430 |       "text/plain": [
 431 |        "  special replace\n",
 432 |        "0                \n",
 433 |        "1       .       .\n",
 434 |        "2       “        \n",
 435 |        "3       \"        \n",
 436 |        "4       -        "
 437 |       ]
 438 |      },
 439 |      "execution_count": 10,
 440 |      "metadata": {},
 441 |      "output_type": "execute_result"
 442 |     }
 443 |    ],
 444 |    "source": [
 445 |     "special_df = pd.read_csv('special.csv', encoding='utf-8')\n",
 446 |     "special_df = special_df.fillna('')\n",
 447 |     "special_df.head()"
 448 |    ]
 449 |   },
 450 |   {
 451 |    "cell_type": "markdown",
 452 |    "metadata": {},
 453 |    "source": [
 454 |     "* special_dict 생성 (special => replace로 변환)"
 455 |    ]
 456 |   },
 457 |   {
 458 |    "cell_type": "code",
 459 |    "execution_count": 11,
 460 |    "metadata": {},
 461 |    "outputs": [],
 462 |    "source": [
 463 |     "special_dict = dict()\n",
 464 |     "\n",
 465 |     "for (special, replace) in zip(special_df['special'], special_df['replace']):\n",
 466 |     "    special_dict[special] = replace\n",
 467 |     "\n",
 468 |     "special_dict[' '] = ' '"
 469 |    ]
 470 |   },
 471 |   {
 472 |    "cell_type": "markdown",
 473 |    "metadata": {},
 474 |    "source": [
 475 |     "* in 연산자로 체크를 위해 list 형식으로 저장"
 476 |    ]
 477 |   },
 478 |   {
 479 |    "cell_type": "code",
 480 |    "execution_count": 12,
 481 |    "metadata": {},
 482 |    "outputs": [],
 483 |    "source": [
 484 |     "specials = list(special_df['special'])\n",
 485 |     "replaces = list(special_df['replace'])"
 486 |    ]
 487 |   },
 488 |   {
 489 |    "cell_type": "markdown",
 490 |    "metadata": {},
 491 |    "source": [
 492 |     "* 특수문자들을 원하는 발음전사로 변환"
 493 |    ]
 494 |   },
 495 |   {
 496 |    "cell_type": "code",
 497 |    "execution_count": 13,
 498 |    "metadata": {},
 499 |    "outputs": [],
 500 |    "source": [
 501 |     "new_corpus = list()\n",
 502 |     "\n",
 503 |     "for sentence in kor_corpus:\n",
 504 |     "    new_sentence = str()\n",
 505 |     "    for ch in sentence:\n",
 506 |     "        if ch in specials:\n",
 507 |     "            new_sentence += special_dict[ch]\n",
 508 |     "        else:\n",
 509 |     "            new_sentence += ch\n",
 510 |     "            \n",
 511 |     "    new_corpus.append(new_sentence)"
 512 |    ]
 513 |   },
 514 |   {
 515 |    "cell_type": "markdown",
 516 |    "metadata": {},
 517 |    "source": [
 518 |     "* 특수문자가 적절히 변경되었는지 확인"
 519 |    ]
 520 |   },
 521 |   {
 522 |    "cell_type": "code",
 523 |    "execution_count": 14,
 524 |    "metadata": {},
 525 |    "outputs": [
 526 |     {
 527 |      "data": {
 528 |       "text/plain": [
 529 |        "['어디 보자...',\n",
 530 |        " '7대 왕국 종족 중에 오크는 처음 듣는다',\n",
 531 |        " '이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.',\n",
 532 |        " '라니스터 가의',\n",
 533 |        " '별 희한한 생각이 다 떠오르곤 하죠',\n",
 534 |        " '아!',\n",
 535 |        " '회의 전에 마실 것 좀 드릴까요? 커피와 차가 있습니다.',\n",
 536 |        " '나는 그런 일에는 전혀 흥미가 없어요.',\n",
 537 |        " '일자리가 있고 제대로 된 회사가 있다면 늘 사람들이 붐비고 활력이 넘치는 도시로 자리매김할 수 있다.',\n",
 538 |        " '카카오가 많이 함유된 다크 초콜릿이 더 좋다는 것을 잊지 말자.',\n",
 539 |        " '비오 13세는 누구인가?',\n",
 540 |        " '네 어..',\n",
 541 |        " '치밀한 사람이긴 하지만...',\n",
 542 |        " '계속 절 따라오셨죠',\n",
 543 |        " '세계 4위 특허출원 강국이지만 지식재산 심사 품질과 보호 수준이 낮아 가치를 제대로 인정받지 못하고 있다.',\n",
 544 |        " '나는 전부터 이 학교를 잘 알고 있었어.',\n",
 545 |        " '뭔가 꿍꿍이를 꾸미는게 틀림없어',\n",
 546 |        " '어쩔 수 없이 과학 수업에 늦었습니다.',\n",
 547 |        " '가르시아 에르난데스 토니 그란데 하비에르 미냐노 등 축구대표팀의 외국인 코치들이 8일 오후  오스트리아 레오강의 대표팀 숙소에서 인터뷰에 응하고있다.',\n",
 548 |        " '검찰은  다 녹음 됐으니까  따지려면 따져볼 수 있다고 말하기도 했다.']"
 549 |       ]
 550 |      },
 551 |      "execution_count": 14,
 552 |      "metadata": {},
 553 |      "output_type": "execute_result"
 554 |     }
 555 |    ],
 556 |    "source": [
 557 |     "new_corpus[:20]"
 558 |    ]
 559 |   },
 560 |   {
 561 |    "cell_type": "markdown",
 562 |    "metadata": {},
 563 |    "source": [
 564 |     "* corpus 업데이트"
 565 |    ]
 566 |   },
 567 |   {
 568 |    "cell_type": "code",
 569 |    "execution_count": 15,
 570 |    "metadata": {},
 571 |    "outputs": [],
 572 |    "source": [
 573 |     "kor_corpus = new_corpus"
 574 |    ]
 575 |   },
 576 |   {
 577 |    "cell_type": "markdown",
 578 |    "metadata": {},
 579 |    "source": [
 580 |     "* 알파벳 필터링 (소문자)"
 581 |    ]
 582 |   },
 583 |   {
 584 |    "cell_type": "code",
 585 |    "execution_count": 16,
 586 |    "metadata": {},
 587 |    "outputs": [],
 588 |    "source": [
 589 |     "alphabet = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']\n",
 590 |     "new_corpus = list()\n",
 591 |     "\n",
 592 |     "for sentence in kor_corpus:\n",
 593 |     "    new_sentence = str()\n",
 594 |     "    for ch in sentence:\n",
 595 |     "        if ch in alphabet:\n",
 596 |     "            continue\n",
 597 |     "        else:\n",
 598 |     "            new_sentence += ch\n",
 599 |     "            \n",
 600 |     "    new_corpus.append(new_sentence)"
 601 |    ]
 602 |   },
 603 |   {
 604 |    "cell_type": "code",
 605 |    "execution_count": 17,
 606 |    "metadata": {},
 607 |    "outputs": [
 608 |     {
 609 |      "data": {
 610 |       "text/plain": [
 611 |        "['어디 보자...',\n",
 612 |        " '7대 왕국 종족 중에 오크는 처음 듣는다',\n",
 613 |        " '이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.',\n",
 614 |        " '라니스터 가의',\n",
 615 |        " '별 희한한 생각이 다 떠오르곤 하죠',\n",
 616 |        " '아!',\n",
 617 |        " '회의 전에 마실 것 좀 드릴까요? 커피와 차가 있습니다.',\n",
 618 |        " '나는 그런 일에는 전혀 흥미가 없어요.',\n",
 619 |        " '일자리가 있고 제대로 된 회사가 있다면 늘 사람들이 붐비고 활력이 넘치는 도시로 자리매김할 수 있다.',\n",
 620 |        " '카카오가 많이 함유된 다크 초콜릿이 더 좋다는 것을 잊지 말자.',\n",
 621 |        " '비오 13세는 누구인가?',\n",
 622 |        " '네 어..',\n",
 623 |        " '치밀한 사람이긴 하지만...',\n",
 624 |        " '계속 절 따라오셨죠',\n",
 625 |        " '세계 4위 특허출원 강국이지만 지식재산 심사 품질과 보호 수준이 낮아 가치를 제대로 인정받지 못하고 있다.',\n",
 626 |        " '나는 전부터 이 학교를 잘 알고 있었어.',\n",
 627 |        " '뭔가 꿍꿍이를 꾸미는게 틀림없어',\n",
 628 |        " '어쩔 수 없이 과학 수업에 늦었습니다.',\n",
 629 |        " '가르시아 에르난데스 토니 그란데 하비에르 미냐노 등 축구대표팀의 외국인 코치들이 8일 오후  오스트리아 레오강의 대표팀 숙소에서 인터뷰에 응하고있다.',\n",
 630 |        " '검찰은  다 녹음 됐으니까  따지려면 따져볼 수 있다고 말하기도 했다.']"
 631 |       ]
 632 |      },
 633 |      "execution_count": 17,
 634 |      "metadata": {},
 635 |      "output_type": "execute_result"
 636 |     }
 637 |    ],
 638 |    "source": [
 639 |     "kor_corpus = new_corpus\n",
 640 |     "kor_corpus[:20]"
 641 |    ]
 642 |   },
 643 |   {
 644 |    "cell_type": "markdown",
 645 |    "metadata": {},
 646 |    "source": [
 647 |     "* 대문자 필터링 (발음으로 전사 ex) K리그 => 케이리그)"
 648 |    ]
 649 |   },
 650 |   {
 651 |    "cell_type": "code",
 652 |    "execution_count": 18,
 653 |    "metadata": {},
 654 |    "outputs": [],
 655 |    "source": [
 656 |     "new_corpus = list()\n",
 657 |     "upper = ['A', 'B', 'C','D','E','F','G','H','I','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z']\n",
 658 |     "upper_dict = {'A' : '에이', 'B' : '비', 'C' : '씨','D' : '디','E' : '이','F' : '에프','G' : '쥐',\n",
 659 |     "              'H' : '에이취','I' : '아이','J' : '제이','K' : '케이','L' : '엘','M' : '엠','N' : '엔',\n",
 660 |     "              'O' : '오','P' : '피','Q' : '큐','R' : '알','S' : '에스','T' : '티','U' : '유','V' : '브이','W' : '떠블유',\n",
 661 |     "              'X' : '엑스','Y' : '와이','Z' : '지'}\n",
 662 |     "\n",
 663 |     "for sentence in kor_corpus:\n",
 664 |     "    new_sentence = str()\n",
 665 |     "    for idx, ch in enumerate(sentence):\n",
 666 |     "        if ch in upper:\n",
 667 |     "            if idx == 0 and (idx == len(sentence) - 1 or sentence[idx + 1] == ' '):\n",
 668 |     "                continue\n",
 669 |     "                \n",
 670 |     "            elif idx != 0 and idx < len(sentence) - 1 and sentence[idx + 1] == ' ' and sentence[idx - 1] == ' ':\n",
 671 |     "                continue\n",
 672 |     "                \n",
 673 |     "            elif idx != 0 and sentence[idx -1] == ' ' and idx + 1 == len(sentence):\n",
 674 |     "                continue\n",
 675 |     "                \n",
 676 |     "            else:\n",
 677 |     "                new_sentence += upper_dict[ch]\n",
 678 |     "            \n",
 679 |     "        else:\n",
 680 |     "            new_sentence += ch\n",
 681 |     "            \n",
 682 |     "    new_corpus.append(new_sentence)"
 683 |    ]
 684 |   },
 685 |   {
 686 |    "cell_type": "code",
 687 |    "execution_count": 19,
 688 |    "metadata": {},
 689 |    "outputs": [
 690 |     {
 691 |      "data": {
 692 |       "text/plain": [
 693 |        "['어디 보자...',\n",
 694 |        " '7대 왕국 종족 중에 오크는 처음 듣는다',\n",
 695 |        " '이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.',\n",
 696 |        " '라니스터 가의',\n",
 697 |        " '별 희한한 생각이 다 떠오르곤 하죠',\n",
 698 |        " '아!',\n",
 699 |        " '회의 전에 마실 것 좀 드릴까요? 커피와 차가 있습니다.',\n",
 700 |        " '나는 그런 일에는 전혀 흥미가 없어요.',\n",
 701 |        " '일자리가 있고 제대로 된 회사가 있다면 늘 사람들이 붐비고 활력이 넘치는 도시로 자리매김할 수 있다.',\n",
 702 |        " '카카오가 많이 함유된 다크 초콜릿이 더 좋다는 것을 잊지 말자.',\n",
 703 |        " '비오 13세는 누구인가?',\n",
 704 |        " '네 어..',\n",
 705 |        " '치밀한 사람이긴 하지만...',\n",
 706 |        " '계속 절 따라오셨죠',\n",
 707 |        " '세계 4위 특허출원 강국이지만 지식재산 심사 품질과 보호 수준이 낮아 가치를 제대로 인정받지 못하고 있다.',\n",
 708 |        " '나는 전부터 이 학교를 잘 알고 있었어.',\n",
 709 |        " '뭔가 꿍꿍이를 꾸미는게 틀림없어',\n",
 710 |        " '어쩔 수 없이 과학 수업에 늦었습니다.',\n",
 711 |        " '가르시아 에르난데스 토니 그란데 하비에르 미냐노 등 축구대표팀의 외국인 코치들이 8일 오후  오스트리아 레오강의 대표팀 숙소에서 인터뷰에 응하고있다.',\n",
 712 |        " '검찰은  다 녹음 됐으니까  따지려면 따져볼 수 있다고 말하기도 했다.']"
 713 |       ]
 714 |      },
 715 |      "execution_count": 19,
 716 |      "metadata": {},
 717 |      "output_type": "execute_result"
 718 |     }
 719 |    ],
 720 |    "source": [
 721 |     "new_corpus[:20]"
 722 |    ]
 723 |   },
 724 |   {
 725 |    "cell_type": "markdown",
 726 |    "metadata": {},
 727 |    "source": [
 728 |     "* 데이터셋 업데이트"
 729 |    ]
 730 |   },
 731 |   {
 732 |    "cell_type": "code",
 733 |    "execution_count": 20,
 734 |    "metadata": {},
 735 |    "outputs": [],
 736 |    "source": [
 737 |     "kor_corpus = new_corpus"
 738 |    ]
 739 |   },
 740 |   {
 741 |    "cell_type": "markdown",
 742 |    "metadata": {},
 743 |    "source": [
 744 |     "* 긴 공백 => ' '"
 745 |    ]
 746 |   },
 747 |   {
 748 |    "cell_type": "code",
 749 |    "execution_count": 21,
 750 |    "metadata": {},
 751 |    "outputs": [],
 752 |    "source": [
 753 |     "new_corpus = list()\n",
 754 |     "\n",
 755 |     "for sentence in kor_corpus:\n",
 756 |     "    new_sentence = str()\n",
 757 |     "    tokens = sentence.split()\n",
 758 |     "    for idx, token in enumerate(tokens):\n",
 759 |     "        if idx == 0:\n",
 760 |     "            new_sentence += token + ' '\n",
 761 |     "        elif idx == len(tokens) - 1:\n",
 762 |     "            new_sentence += token\n",
 763 |     "        else:\n",
 764 |     "            new_sentence += token + ' '\n",
 765 |     "            \n",
 766 |     "    new_corpus.append(new_sentence)"
 767 |    ]
 768 |   },
 769 |   {
 770 |    "cell_type": "code",
 771 |    "execution_count": 22,
 772 |    "metadata": {},
 773 |    "outputs": [
 774 |     {
 775 |      "data": {
 776 |       "text/plain": [
 777 |        "['어디 보자...',\n",
 778 |        " '7대 왕국 종족 중에 오크는 처음 듣는다',\n",
 779 |        " '이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.',\n",
 780 |        " '라니스터 가의',\n",
 781 |        " '별 희한한 생각이 다 떠오르곤 하죠',\n",
 782 |        " '아! ',\n",
 783 |        " '회의 전에 마실 것 좀 드릴까요? 커피와 차가 있습니다.',\n",
 784 |        " '나는 그런 일에는 전혀 흥미가 없어요.',\n",
 785 |        " '일자리가 있고 제대로 된 회사가 있다면 늘 사람들이 붐비고 활력이 넘치는 도시로 자리매김할 수 있다.',\n",
 786 |        " '카카오가 많이 함유된 다크 초콜릿이 더 좋다는 것을 잊지 말자.',\n",
 787 |        " '비오 13세는 누구인가?',\n",
 788 |        " '네 어..',\n",
 789 |        " '치밀한 사람이긴 하지만...',\n",
 790 |        " '계속 절 따라오셨죠',\n",
 791 |        " '세계 4위 특허출원 강국이지만 지식재산 심사 품질과 보호 수준이 낮아 가치를 제대로 인정받지 못하고 있다.',\n",
 792 |        " '나는 전부터 이 학교를 잘 알고 있었어.',\n",
 793 |        " '뭔가 꿍꿍이를 꾸미는게 틀림없어',\n",
 794 |        " '어쩔 수 없이 과학 수업에 늦었습니다.',\n",
 795 |        " '가르시아 에르난데스 토니 그란데 하비에르 미냐노 등 축구대표팀의 외국인 코치들이 8일 오후 오스트리아 레오강의 대표팀 숙소에서 인터뷰에 응하고있다.',\n",
 796 |        " '검찰은 다 녹음 됐으니까 따지려면 따져볼 수 있다고 말하기도 했다.']"
 797 |       ]
 798 |      },
 799 |      "execution_count": 22,
 800 |      "metadata": {},
 801 |      "output_type": "execute_result"
 802 |     }
 803 |    ],
 804 |    "source": [
 805 |     "new_corpus[:20]"
 806 |    ]
 807 |   },
 808 |   {
 809 |    "cell_type": "markdown",
 810 |    "metadata": {},
 811 |    "source": [
 812 |     "* 데이터셋 업데이트"
 813 |    ]
 814 |   },
 815 |   {
 816 |    "cell_type": "code",
 817 |    "execution_count": 23,
 818 |    "metadata": {},
 819 |    "outputs": [],
 820 |    "source": [
 821 |     "kor_corpus = new_corpus"
 822 |    ]
 823 |   },
 824 |   {
 825 |    "cell_type": "markdown",
 826 |    "metadata": {},
 827 |    "source": [
 828 |     "* 데이터셋 크기 확인"
 829 |    ]
 830 |   },
 831 |   {
 832 |    "cell_type": "code",
 833 |    "execution_count": 24,
 834 |    "metadata": {},
 835 |    "outputs": [
 836 |     {
 837 |      "data": {
 838 |       "text/plain": [
 839 |        "3048264"
 840 |       ]
 841 |      },
 842 |      "execution_count": 24,
 843 |      "metadata": {},
 844 |      "output_type": "execute_result"
 845 |     }
 846 |    ],
 847 |    "source": [
 848 |     "len(kor_corpus)"
 849 |    ]
 850 |   },
 851 |   {
 852 |    "cell_type": "markdown",
 853 |    "metadata": {},
 854 |    "source": [
 855 |     "* 필터링 되면서 빈문자열만 남은 데이터 제외"
 856 |    ]
 857 |   },
 858 |   {
 859 |    "cell_type": "code",
 860 |    "execution_count": 25,
 861 |    "metadata": {},
 862 |    "outputs": [],
 863 |    "source": [
 864 |     "new_corpus = list()\n",
 865 |     "\n",
 866 |     "for sentence in kor_corpus:\n",
 867 |     "    if len(sentence) == 0:\n",
 868 |     "        continue\n",
 869 |     "    new_corpus.append(sentence)"
 870 |    ]
 871 |   },
 872 |   {
 873 |    "cell_type": "markdown",
 874 |    "metadata": {},
 875 |    "source": [
 876 |     "* 약 25,000개 필터링"
 877 |    ]
 878 |   },
 879 |   {
 880 |    "cell_type": "code",
 881 |    "execution_count": 26,
 882 |    "metadata": {},
 883 |    "outputs": [
 884 |     {
 885 |      "data": {
 886 |       "text/plain": [
 887 |        "3023296"
 888 |       ]
 889 |      },
 890 |      "execution_count": 26,
 891 |      "metadata": {},
 892 |      "output_type": "execute_result"
 893 |     }
 894 |    ],
 895 |    "source": [
 896 |     "len(new_corpus)"
 897 |    ]
 898 |   },
 899 |   {
 900 |    "cell_type": "markdown",
 901 |    "metadata": {},
 902 |    "source": [
 903 |     "* 데이터셋 업데이트"
 904 |    ]
 905 |   },
 906 |   {
 907 |    "cell_type": "code",
 908 |    "execution_count": 27,
 909 |    "metadata": {},
 910 |    "outputs": [],
 911 |    "source": [
 912 |     "kor_corpus = new_corpus"
 913 |    ]
 914 |   },
 915 |   {
 916 |    "cell_type": "markdown",
 917 |    "metadata": {},
 918 |    "source": [
 919 |     "* 아무지점이나 찍어서 중간확인"
 920 |    ]
 921 |   },
 922 |   {
 923 |    "cell_type": "code",
 924 |    "execution_count": 28,
 925 |    "metadata": {},
 926 |    "outputs": [
 927 |     {
 928 |      "data": {
 929 |       "text/plain": [
 930 |        "['이란 세계식량계획 사무소 대표 및 국가 담당국장 쥐는 2018년 올해 주이란 한국대사관 측이 220만 달러를 기부했을 때 한국 국민과 정부에 감사의 뜻을 전하기도 했다.',\n",
 931 |        " '조심해요 오염된 거예요',\n",
 932 |        " '그건 그렇고 메이즈는 어때?',\n",
 933 |        " '간에 있는 덩어리가..',\n",
 934 |        " '. ',\n",
 935 |        " '물렸어요? ',\n",
 936 |        " '안 보이나 보네',\n",
 937 |        " '그 전엔 아무것도 쓸 것 이 없었어요. 닥터 그레이.',\n",
 938 |        " '하루만 머무를 거요',\n",
 939 |        " '그런건 내 일이 아니라서 다른 도움이 필요하면 말해 난 쇼파에 있을꺼야',\n",
 940 |        " '고마워 ',\n",
 941 |        " '취업준비생들은 현실적으로 청년취업이 어려운 상황에서 아쉽다는 반응이었다.',\n",
 942 |        " '시장을 떠나는 순간 덮치겠음',\n",
 943 |        " '그리고 디자인이랑',\n",
 944 |        " '저항권이란 위헌적인 국가권력의 행사 때문에 헌법이 침해 또는 파괴되는 경우 주권자인 국민이 헌법을 보장하기 위해 최후 비상수단으로 국가권력에 실력으로 저항하는 것입니다.',\n",
 945 |        " '현대모비스 양동근도 정규 리그에서 상대해 본 결과 함지훈을 막을 선수가 없다는 게 전자랜드의 약점이라고 지적했다.',\n",
 946 |        " '아사히는 꾸준히 한국어 공부를 하면서 내가 하고 싶은 노래를 만들기 위해 항상 들고 다닌다고 말해 아티스트로서 더욱 발전할 수 있는 가능성을 보여줬다.',\n",
 947 |        " '서양마사지 방식으로는 스웨디시 마사지 에스알엔 그리고 홀리스틱 마사지가 있다.',\n",
 948 |        " '내 말을 들었더라면',\n",
 949 |        " '앞으로 이렇게 수도권 광역 경제발전위원회가 자주 열려서 다양한 이야기를 해야 한다.']"
 950 |       ]
 951 |      },
 952 |      "execution_count": 28,
 953 |      "metadata": {},
 954 |      "output_type": "execute_result"
 955 |     }
 956 |    ],
 957 |    "source": [
 958 |     "kor_corpus[6000:6020]"
 959 |    ]
 960 |   },
 961 |   {
 962 |    "cell_type": "markdown",
 963 |    "metadata": {},
 964 |    "source": [
 965 |     "* 뒤에가 공백으로 끝나는 문자열 공백 제거"
 966 |    ]
 967 |   },
 968 |   {
 969 |    "cell_type": "code",
 970 |    "execution_count": 29,
 971 |    "metadata": {},
 972 |    "outputs": [],
 973 |    "source": [
 974 |     "for idx, sentence in enumerate(kor_corpus):\n",
 975 |     "    if sentence[-1] == ' ':\n",
 976 |     "        kor_corpus[idx] = sentence[:-1] "
 977 |    ]
 978 |   },
 979 |   {
 980 |    "cell_type": "code",
 981 |    "execution_count": 30,
 982 |    "metadata": {},
 983 |    "outputs": [
 984 |     {
 985 |      "data": {
 986 |       "text/plain": [
 987 |        "['이란 세계식량계획 사무소 대표 및 국가 담당국장 쥐는 2018년 올해 주이란 한국대사관 측이 220만 달러를 기부했을 때 한국 국민과 정부에 감사의 뜻을 전하기도 했다.',\n",
 988 |        " '조심해요 오염된 거예요',\n",
 989 |        " '그건 그렇고 메이즈는 어때?',\n",
 990 |        " '간에 있는 덩어리가..',\n",
 991 |        " '.',\n",
 992 |        " '물렸어요?',\n",
 993 |        " '안 보이나 보네',\n",
 994 |        " '그 전엔 아무것도 쓸 것 이 없었어요. 닥터 그레이.',\n",
 995 |        " '하루만 머무를 거요',\n",
 996 |        " '그런건 내 일이 아니라서 다른 도움이 필요하면 말해 난 쇼파에 있을꺼야',\n",
 997 |        " '고마워',\n",
 998 |        " '취업준비생들은 현실적으로 청년취업이 어려운 상황에서 아쉽다는 반응이었다.',\n",
 999 |        " '시장을 떠나는 순간 덮치겠음',\n",
1000 |        " '그리고 디자인이랑',\n",
1001 |        " '저항권이란 위헌적인 국가권력의 행사 때문에 헌법이 침해 또는 파괴되는 경우 주권자인 국민이 헌법을 보장하기 위해 최후 비상수단으로 국가권력에 실력으로 저항하는 것입니다.',\n",
1002 |        " '현대모비스 양동근도 정규 리그에서 상대해 본 결과 함지훈을 막을 선수가 없다는 게 전자랜드의 약점이라고 지적했다.',\n",
1003 |        " '아사히는 꾸준히 한국어 공부를 하면서 내가 하고 싶은 노래를 만들기 위해 항상 들고 다닌다고 말해 아티스트로서 더욱 발전할 수 있는 가능성을 보여줬다.',\n",
1004 |        " '서양마사지 방식으로는 스웨디시 마사지 에스알엔 그리고 홀리스틱 마사지가 있다.',\n",
1005 |        " '내 말을 들었더라면',\n",
1006 |        " '앞으로 이렇게 수도권 광역 경제발전위원회가 자주 열려서 다양한 이야기를 해야 한다.']"
1007 |       ]
1008 |      },
1009 |      "execution_count": 30,
1010 |      "metadata": {},
1011 |      "output_type": "execute_result"
1012 |     }
1013 |    ],
1014 |    "source": [
1015 |     "kor_corpus[6000:6020]"
1016 |    ]
1017 |   },
1018 |   {
1019 |    "cell_type": "markdown",
1020 |    "metadata": {},
1021 |    "source": [
1022 |     "* 현재 데이터셋 크기 확인"
1023 |    ]
1024 |   },
1025 |   {
1026 |    "cell_type": "code",
1027 |    "execution_count": 31,
1028 |    "metadata": {},
1029 |    "outputs": [
1030 |     {
1031 |      "data": {
1032 |       "text/plain": [
1033 |        "3023296"
1034 |       ]
1035 |      },
1036 |      "execution_count": 31,
1037 |      "metadata": {},
1038 |      "output_type": "execute_result"
1039 |     }
1040 |    ],
1041 |    "source": [
1042 |     "len(kor_corpus)"
1043 |    ]
1044 |   },
1045 |   {
1046 |    "cell_type": "markdown",
1047 |    "metadata": {},
1048 |    "source": [
1049 |     "* '?', '!', '.'만 남은 문자열 필터링"
1050 |    ]
1051 |   },
1052 |   {
1053 |    "cell_type": "code",
1054 |    "execution_count": 32,
1055 |    "metadata": {},
1056 |    "outputs": [
1057 |     {
1058 |      "data": {
1059 |       "text/plain": [
1060 |        "3001447"
1061 |       ]
1062 |      },
1063 |      "execution_count": 32,
1064 |      "metadata": {},
1065 |      "output_type": "execute_result"
1066 |     }
1067 |    ],
1068 |    "source": [
1069 |     "new_corpus = list()\n",
1070 |     "\n",
1071 |     "for sentence in kor_corpus:\n",
1072 |     "    if sentence == '?' or sentence == '!' or sentence == '.':\n",
1073 |     "        continue\n",
1074 |     "        \n",
1075 |     "    else:\n",
1076 |     "        new_corpus.append(sentence)\n",
1077 |     "        \n",
1078 |     "len(new_corpus)"
1079 |    ]
1080 |   },
1081 |   {
1082 |    "cell_type": "markdown",
1083 |    "metadata": {},
1084 |    "source": [
1085 |     "* 약 2만개 필터링 후 데이터셋 업데이트"
1086 |    ]
1087 |   },
1088 |   {
1089 |    "cell_type": "code",
1090 |    "execution_count": 33,
1091 |    "metadata": {},
1092 |    "outputs": [],
1093 |    "source": [
1094 |     "kor_corpus = new_corpus"
1095 |    ]
1096 |   },
1097 |   {
1098 |    "cell_type": "markdown",
1099 |    "metadata": {},
1100 |    "source": [
1101 |     "* import **KoNPron** (Korean Number Pronunciaion) => 숫자를 발음전사로 변환"
1102 |    ]
1103 |   },
1104 |   {
1105 |    "cell_type": "code",
1106 |    "execution_count": 34,
1107 |    "metadata": {},
1108 |    "outputs": [],
1109 |    "source": [
1110 |     "from konpron import KoNPron\n",
1111 |     "\n",
1112 |     "kpr = KoNPron()"
1113 |    ]
1114 |   },
1115 |   {
1116 |    "cell_type": "markdown",
1117 |    "metadata": {},
1118 |    "source": [
1119 |     "* **KoNPron** 테스트"
1120 |    ]
1121 |   },
1122 |   {
1123 |    "cell_type": "code",
1124 |    "execution_count": 35,
1125 |    "metadata": {},
1126 |    "outputs": [
1127 |     {
1128 |      "name": "stdout",
1129 |      "output_type": "stream",
1130 |      "text": [
1131 |       "천이백삼십사랑, 십이 점 삼사랑, 천이백삼십사랑, 백이십삼의 사 승랑, 일의 이백삼십사 승랑, 일 점 이삼의 사 승랑, 십이 점 삼의 사 승랑, 일이삼사랑, 이억 구천삼백삼십만 천삼백 점 오오삼영일의 사 승\n"
1132 |      ]
1133 |     }
1134 |    ],
1135 |    "source": [
1136 |     "sentence = \"1234랑, 12.34랑, 1,234랑, 123^4랑, 1²³⁴랑, 1.23^4랑, 12.3⁴랑, 12·34랑, 293301300.55301⁴\"\n",
1137 |     "\n",
1138 |     "ret = kpr.convert(sentence)\n",
1139 |     "if ret is not None:\n",
1140 |     "    print(ret)"
1141 |    ]
1142 |   },
1143 |   {
1144 |    "cell_type": "markdown",
1145 |    "metadata": {},
1146 |    "source": [
1147 |     "* **KoNPron**으로 숫자 => 발음전사로 변환"
1148 |    ]
1149 |   },
1150 |   {
1151 |    "cell_type": "code",
1152 |    "execution_count": 36,
1153 |    "metadata": {},
1154 |    "outputs": [],
1155 |    "source": [
1156 |     "new_corpus = list()\n",
1157 |     "\n",
1158 |     "for sentence in kor_corpus:\n",
1159 |     "    try:\n",
1160 |     "        ret = kpr.convert(sentence)\n",
1161 |     "        if ret is not None:\n",
1162 |     "            new_corpus.append(ret)\n",
1163 |     "        else:\n",
1164 |     "            new_corpus.append(sentence)\n",
1165 |     "    except:\n",
1166 |     "        print(sentence)"
1167 |    ]
1168 |   },
1169 |   {
1170 |    "cell_type": "code",
1171 |    "execution_count": 37,
1172 |    "metadata": {},
1173 |    "outputs": [],
1174 |    "source": [
1175 |     "kor_corpus = new_corpus"
1176 |    ]
1177 |   },
1178 |   {
1179 |    "cell_type": "markdown",
1180 |    "metadata": {},
1181 |    "source": [
1182 |     "* 음향모델에서 사용하는 2,040개의 문자 레이블로만 이루어진 문장 수 확인"
1183 |    ]
1184 |   },
1185 |   {
1186 |    "cell_type": "code",
1187 |    "execution_count": 38,
1188 |    "metadata": {},
1189 |    "outputs": [
1190 |     {
1191 |      "name": "stdout",
1192 |      "output_type": "stream",
1193 |      "text": [
1194 |       "2992120\n"
1195 |      ]
1196 |     }
1197 |    ],
1198 |    "source": [
1199 |     "acoustic_labels = list(pd.read_csv('train_labels.csv')['char'])\n",
1200 |     "count = 0\n",
1201 |     "\n",
1202 |     "for sentence in kor_corpus:\n",
1203 |     "    for ch in sentence:\n",
1204 |     "        if ch not in acoustic_labels:\n",
1205 |     "            count += 1\n",
1206 |     "            break\n",
1207 |     "\n",
1208 |     "print(len(kor_corpus) - count)"
1209 |    ]
1210 |   },
1211 |   {
1212 |    "cell_type": "markdown",
1213 |    "metadata": {},
1214 |    "source": [
1215 |     "* 조건에 만족하는 2,992,638개의 문장들로 최종 코퍼스 구성"
1216 |    ]
1217 |   },
1218 |   {
1219 |    "cell_type": "code",
1220 |    "execution_count": 41,
1221 |    "metadata": {},
1222 |    "outputs": [
1223 |     {
1224 |      "data": {
1225 |       "text/plain": [
1226 |        "2992120"
1227 |       ]
1228 |      },
1229 |      "execution_count": 41,
1230 |      "metadata": {},
1231 |      "output_type": "execute_result"
1232 |     }
1233 |    ],
1234 |    "source": [
1235 |     "final_corpus = list()\n",
1236 |     "\n",
1237 |     "for sentence in kor_corpus:\n",
1238 |     "    flag = True\n",
1239 |     "    for ch in sentence:\n",
1240 |     "        if ch not in acoustic_labels:\n",
1241 |     "            flag = False\n",
1242 |     "            break\n",
1243 |     "    if flag:\n",
1244 |     "        final_corpus.append(sentence)\n",
1245 |     "\n",
1246 |     "len(final_corpus)"
1247 |    ]
1248 |   },
1249 |   {
1250 |    "cell_type": "code",
1251 |    "execution_count": 44,
1252 |    "metadata": {},
1253 |    "outputs": [],
1254 |    "source": [
1255 |     "import csv\n",
1256 |     "\n",
1257 |     "def load_label(label_path, encoding='utf-8'):\n",
1258 |     "    char2id = dict()\n",
1259 |     "    id2char = dict()\n",
1260 |     "\n",
1261 |     "    with open(label_path, 'r', encoding=encoding) as f:\n",
1262 |     "        labels = csv.reader(f, delimiter=',')\n",
1263 |     "        next(labels)\n",
1264 |     "\n",
1265 |     "        for row in labels:\n",
1266 |     "            char2id[row[1]] = row[0]\n",
1267 |     "            id2char[int(row[0])] = row[1]\n",
1268 |     "\n",
1269 |     "    return char2id, id2char\n",
1270 |     "\n",
1271 |     "char2id, id2char = load_label('train_labels.csv')"
1272 |    ]
1273 |   },
1274 |   {
1275 |    "cell_type": "code",
1276 |    "execution_count": 49,
1277 |    "metadata": {},
1278 |    "outputs": [
1279 |     {
1280 |      "name": "stdout",
1281 |      "output_type": "stream",
1282 |      "text": [
1283 |       "2992120\n",
1284 |       "2992120\n"
1285 |      ]
1286 |     }
1287 |    ],
1288 |    "source": [
1289 |     "corpus_labels = list()\n",
1290 |     "\n",
1291 |     "for sentence in final_corpus:\n",
1292 |     "    label = str()\n",
1293 |     "    for ch in sentence:\n",
1294 |     "        if idx == 0:\n",
1295 |     "            label += char2id[ch] + ' '\n",
1296 |     "        elif idx == len(sentence) - 1:\n",
1297 |     "            label += char2id[ch]\n",
1298 |     "        else:\n",
1299 |     "            label += char2id[ch] + ' '\n",
1300 |     "    \n",
1301 |     "    corpus_labels.append(label)\n",
1302 |     "    \n",
1303 |     "print(len(final_corpus))\n",
1304 |     "print(len(corpus_labels))"
1305 |    ]
1306 |   },
1307 |   {
1308 |    "cell_type": "code",
1309 |    "execution_count": 50,
1310 |    "metadata": {},
1311 |    "outputs": [
1312 |     {
1313 |      "data": {
1314 |       "text/html": [
1315 |        "<div>\n",
1316 |        "<style scoped>\n",
1317 |        "    .dataframe tbody tr th:only-of-type {\n",
1318 |        "        vertical-align: middle;\n",
1319 |        "    }\n",
1320 |        "\n",
1321 |        "    .dataframe tbody tr th {\n",
1322 |        "        vertical-align: top;\n",
1323 |        "    }\n",
1324 |        "\n",
1325 |        "    .dataframe thead th {\n",
1326 |        "        text-align: right;\n",
1327 |        "    }\n",
1328 |        "</style>\n",
1329 |        "<table border=\"1\" class=\"dataframe\">\n",
1330 |        "  <thead>\n",
1331 |        "    <tr style=\"text-align: right;\">\n",
1332 |        "      <th></th>\n",
1333 |        "      <th>ko</th>\n",
1334 |        "      <th>id</th>\n",
1335 |        "    </tr>\n",
1336 |        "  </thead>\n",
1337 |        "  <tbody>\n",
1338 |        "    <tr>\n",
1339 |        "      <th>0</th>\n",
1340 |        "      <td>어디 보자...</td>\n",
1341 |        "      <td>8 190 0 42 45 1 1 1</td>\n",
1342 |        "    </tr>\n",
1343 |        "    <tr>\n",
1344 |        "      <th>1</th>\n",
1345 |        "      <td>칠대 왕국 종족 중에 오크는 처음 듣는다</td>\n",
1346 |        "      <td>318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...</td>\n",
1347 |        "    </tr>\n",
1348 |        "    <tr>\n",
1349 |        "      <th>2</th>\n",
1350 |        "      <td>이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.</td>\n",
1351 |        "      <td>3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...</td>\n",
1352 |        "    </tr>\n",
1353 |        "    <tr>\n",
1354 |        "      <th>3</th>\n",
1355 |        "      <td>라니스터 가의</td>\n",
1356 |        "      <td>32 20 79 162 0 6 130</td>\n",
1357 |        "    </tr>\n",
1358 |        "    <tr>\n",
1359 |        "      <th>4</th>\n",
1360 |        "      <td>별 희한한 생각이 다 떠오르곤 하죠</td>\n",
1361 |        "      <td>233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...</td>\n",
1362 |        "    </tr>\n",
1363 |        "  </tbody>\n",
1364 |        "</table>\n",
1365 |        "</div>"
1366 |       ],
1367 |       "text/plain": [
1368 |        "                                           ko  \\\n",
1369 |        "0                                    어디 보자...   \n",
1370 |        "1                      칠대 왕국 종족 중에 오크는 처음 듣는다   \n",
1371 |        "2  이번에 새로 개시할 제품 계열은 광고 모델을 계약해서 홍보하는 게 좋겠어요.   \n",
1372 |        "3                                     라니스터 가의   \n",
1373 |        "4                         별 희한한 생각이 다 떠오르곤 하죠   \n",
1374 |        "\n",
1375 |        "                                                  id  \n",
1376 |        "0                               8 190 0 42 45 1 1 1   \n",
1377 |        "1  318 50 0 576 170 0 363 401 0 129 17 0 57 238 4...  \n",
1378 |        "2  3 93 17 0 260 38 0 168 47 91 0 43 467 0 142 22...  \n",
1379 |        "3                              32 20 79 162 0 6 130   \n",
1380 |        "4  233 0 439 27 27 0 71 100 3 0 15 0 512 57 114 5...  "
1381 |       ]
1382 |      },
1383 |      "execution_count": 50,
1384 |      "metadata": {},
1385 |      "output_type": "execute_result"
1386 |     }
1387 |    ],
1388 |    "source": [
1389 |     "corpus_dict = {'ko' : final_corpus,\n",
1390 |     "               'id' : corpus_labels}\n",
1391 |     "corpus_df = pd.DataFrame(corpus_dict)\n",
1392 |     "\n",
1393 |     "corpus_df.head()"
1394 |    ]
1395 |   },
1396 |   {
1397 |    "cell_type": "code",
1398 |    "execution_count": 53,
1399 |    "metadata": {},
1400 |    "outputs": [
1401 |     {
1402 |      "data": {
1403 |       "text/plain": [
1404 |        "2992120"
1405 |       ]
1406 |      },
1407 |      "execution_count": 53,
1408 |      "metadata": {},
1409 |      "output_type": "execute_result"
1410 |     }
1411 |    ],
1412 |    "source": [
1413 |     "len(corpus_df)"
1414 |    ]
1415 |   },
1416 |   {
1417 |    "cell_type": "markdown",
1418 |    "metadata": {},
1419 |    "source": [
1420 |     "* 데이터셋 피클로 저장"
1421 |    ]
1422 |   },
1423 |   {
1424 |    "cell_type": "code",
1425 |    "execution_count": 52,
1426 |    "metadata": {},
1427 |    "outputs": [],
1428 |    "source": [
1429 |     "import pickle\n",
1430 |     "\n",
1431 |     "with open('corpus_df.bin', 'wb') as f:\n",
1432 |     "    pickle.dump(corpus_df, f)"
1433 |    ]
1434 |   }
1435 |  ],
1436 |  "metadata": {
1437 |   "kernelspec": {
1438 |    "display_name": "Python 3",
1439 |    "language": "python",
1440 |    "name": "python3"
1441 |   },
1442 |   "language_info": {
1443 |    "codemirror_mode": {
1444 |     "name": "ipython",
1445 |     "version": 3
1446 |    },
1447 |    "file_extension": ".py",
1448 |    "mimetype": "text/x-python",
1449 |    "name": "python",
1450 |    "nbconvert_exporter": "python",
1451 |    "pygments_lexer": "ipython3",
1452 |    "version": "3.7.6"
1453 |   }
1454 |  },
1455 |  "nbformat": 4,
1456 |  "nbformat_minor": 2
1457 | }
1458 | 


--------------------------------------------------------------------------------
/preprocess/konpron.py:
--------------------------------------------------------------------------------
  1 | class KoNPron:
  2 |     def __init__(self):
  3 |         self.base_digit = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
  4 |         self.super_digit = ['⁰', '¹', '²', '³', '⁴', '⁵', '⁶', '⁷', '⁸', '⁹']
  5 |         self.small_scale = ['', '십', '백', '천']
  6 |         self.large_scale = ['', '만 ', '억 ', '조 ', '경 ', '해 ']
  7 |         self.literal = ['영', '일', '이', '삼', '사', '오', '육', '칠', '팔', '구']
  8 |         self.spoken_unit = ['', '하나', '둘', '셋', '넷', '다섯', '여섯', '일곱', '여덟', '아홉']
  9 |         self.spoke_tens = ['', '열', '스물', '서른', '마흔', '쉰', '예순', '일흔', '여든', '아흔']
 10 |         self.sentence = str()
 11 | 
 12 |     def _detect(self, sentence):
 13 |         self.sentence = sentence
 14 |         detection_data = list()
 15 |         tmp = str()
 16 | 
 17 |         total_len = len(sentence)
 18 |         point_count = 0
 19 |         continuous_count = 0
 20 |         digit_type = 'vanilla'
 21 | 
 22 |         detected = False
 23 |         zero_started = False
 24 | 
 25 |         for idx, char in enumerate(sentence):
 26 |             if char in self.base_digit:
 27 |                 if not detected:
 28 |                     detected = True
 29 |                     if char is '0':
 30 |                         zero_started = True
 31 |                 tmp += char
 32 |                 continuous_count += 1
 33 |                 if zero_started and continuous_count > 8:
 34 |                     digit_type = 'telephone/none'
 35 |                 if continuous_count > 20:
 36 |                     digit_type = 'enormous/none'
 37 |             else:
 38 |                 continous_count = 0
 39 |                 zero_started = False
 40 |                 if char == ',':
 41 |                     if idx + 1 < total_len and idx > 0:
 42 |                         if sentence[idx - 1] in self.base_digit and sentence[idx + 1] in self.base_digit:
 43 |                             tmp += char
 44 |                 elif char == '.':
 45 |                     if idx + 1 < total_len and idx > 0:
 46 |                         if sentence[idx - 1] in self.base_digit and sentence[idx + 1] in self.base_digit:
 47 |                             point_count += 1
 48 |                             if point_count == 1:
 49 |                                 digit_type = 'fraction'
 50 |                             if point_count > 1:
 51 |                                 digit_type = 'version'
 52 |                             tmp += char
 53 |                 elif char == '^':
 54 |                     if idx + 1 < total_len and idx > 0:
 55 |                         if sentence[idx - 1] in self.base_digit and sentence[idx + 1] in self.base_digit:
 56 |                             digit_type += '/square'
 57 |                             tmp += char
 58 |                         else:
 59 |                             if digit_type != 'exception/none':
 60 |                                 digit_type = 'exception/none'
 61 |                                 tmp += char
 62 |                 elif char in self.super_digit:
 63 |                     if idx > 0:
 64 |                         if sentence[idx - 1] in self.base_digit or sentence[idx - 1] in self.super_digit:
 65 |                             if not 'square' in digit_type:
 66 |                                 digit_type += '/square'
 67 |                             tmp += char
 68 |                         else:
 69 |                             if digit_type != 'exception/none':
 70 |                                 digit_type = 'exception/none'
 71 |                                 tmp += char
 72 |                 elif char == '·':
 73 |                     if idx + 1 < total_len and idx > 0:
 74 |                         if sentence[idx - 1] in self.base_digit and sentence[idx + 1] in self.base_digit:
 75 |                             digit_type = 'date'
 76 |                             tmp += char
 77 |                 else:
 78 |                     if detected:
 79 |                         detected = False
 80 |                         if '/' not in digit_type:
 81 |                             digit_type += '/none'
 82 |                         detection_data.append((digit_type, tmp))
 83 |                         tmp = str()
 84 |                         digit_type = 'vanilla'
 85 |                         point_count = 0
 86 |                         continuous_count = 0
 87 |                     else:
 88 |                         if tmp:
 89 |                             detection_data.append((digit_type, tmp))
 90 |                             tmp = str()
 91 |                             digit_type = 'vanilla'
 92 |                             point_count = 0
 93 |                             continuous_count = 0
 94 | 
 95 |         if detected:
 96 |             if '/' not in digit_type:
 97 |                 digit_type += '/none'
 98 |             detection_data.append((digit_type, tmp))
 99 | 
100 |         elif tmp:
101 |             detection_data.append((digit_type, tmp))
102 |         return detection_data
103 | 
104 |     def _preprocess(self, detection_data):
105 |         preprocessed_data = list()
106 | 
107 |         for digit_type, target in detection_data:
108 |             original = target
109 |             target_seq = list()
110 |             reading_method = list()
111 |             target_len = len(target)
112 | 
113 |             main_type, sub_type = digit_type.split('/')
114 |             if main_type == 'exception':
115 |                 return None
116 | 
117 |             if main_type == 'version':
118 |                 splited = target.split('.')
119 |                 for count, frag in enumerate(splited):
120 |                     target_seq.append(frag)
121 |                     reading_method.append('individual')
122 |                     if count < len(splited) - 1:
123 |                         target_seq.append('.')
124 |                         reading_method.append('point')
125 |                 preprocessed_data.append((reading_method, target_seq, original))
126 | 
127 |             if main_type == 'date':
128 |                 target = target.split('·')
129 |                 for frag in target:
130 |                     target_seq.append(frag)
131 |                     reading_method.append('individual')
132 |                 preprocessed_data.append((reading_method, target_seq, original))
133 | 
134 |             if main_type == 'telephone' or main_type == 'enormous':
135 |                 target = target.replace('0', '_')
136 |                 target_seq.append(target)
137 |                 reading_method.append('individual')
138 |                 preprocessed_data.append((reading_method, target_seq, original))
139 | 
140 |             if sub_type != 'none':
141 |                 if main_type == 'vanilla':
142 |                     target = [target.replace(',', '')]
143 |                 if main_type == 'fraction':
144 |                     target = target.split('.')
145 | 
146 |                 if sub_type == 'square':
147 |                     if '^' not in target[-1]:
148 |                         super_part = str()
149 |                         for idx, digit in enumerate(target[-1]):
150 |                             if digit not in self.base_digit:
151 |                                 super_idx = idx
152 |                                 break
153 |                         super_num = list(target[-1][idx:])
154 |                         for idx in range(len(super_num)):
155 |                             super_num[idx] = str(self.super_digit.index(super_num[idx]))
156 |                         super_num = ''.join(super_num)
157 |                         super_part += target[-1][:super_idx] + '^' + super_num
158 |                         if main_type == 'vanilla':
159 |                             target = [super_part]
160 |                         else:
161 |                             target = target[:1] + [super_part]
162 |                     if main_type == 'fraction':
163 |                         target_seq.append(target[0])
164 |                         reading_method.append('literal')
165 |                         target_seq.append('.')
166 |                         reading_method.append('point')
167 | 
168 |                     tmp_len = len(target_seq)
169 |                     tmp = target[-1].split('^')
170 |                     for seq in tmp:
171 |                         target_seq.append(seq)
172 |                         if 'point' in reading_method:
173 |                             if 'of' in reading_method:
174 |                                 reading_method.append('literal')
175 |                             else:
176 |                                 reading_method.append('individual')
177 |                         else:
178 |                             reading_method.append('literal')
179 |                         if len(target_seq) == tmp_len + 1:
180 |                             target_seq.append("^")
181 |                             reading_method.append("of")
182 |                         elif len(target_seq) == tmp_len + 3:
183 |                             target_seq.append("^")
184 |                             reading_method.append("super")
185 |                     preprocessed_data.append((reading_method, target_seq, original))
186 | 
187 |             else:
188 |                 if main_type == 'vanilla':
189 |                     target = target.replace(',', '')
190 |                     target_seq.append(target)
191 |                     reading_method.append('literal')
192 |                     preprocessed_data.append((reading_method, target_seq, original))
193 | 
194 |                 if main_type == 'fraction':
195 |                     target = target.split('.')
196 |                     for frag in target:
197 |                         target_seq.append(frag)
198 |                         if 'point' in reading_method:
199 |                             reading_method.append('individual')
200 |                         else:
201 |                             reading_method.append('literal')
202 |                         if len(target_seq) == 1:
203 |                             target_seq.append(".")
204 |                             reading_method.append('point')
205 | 
206 |                     preprocessed_data.append((reading_method, target_seq, original))
207 | 
208 |         return preprocessed_data
209 | 
210 |     def _read(self, preprocessed_data, mode='informal'):
211 |         def literal_read(self, frag, mode='informal'):
212 |             korean = str()
213 |             tmp = str()
214 |             length = len(frag)
215 |             for idx, digit in enumerate(frag):
216 |                 digit = int(digit)
217 |                 inversed_idx = length - idx - 1
218 |                 if mode == 'formal':
219 |                     if inversed_idx % 4:
220 |                         if digit:
221 |                             tmp += self.literal[digit]
222 |                             tmp += self.small_scale[inversed_idx % 4]
223 |                     else:
224 |                         if digit or length == 1:
225 |                             tmp += self.literal[digit]
226 |                         if tmp:
227 |                             tmp += self.large_scale[inversed_idx // 4]
228 |                         korean += tmp
229 |                         tmp = str()
230 | 
231 |                 elif mode == 'informal':
232 |                     if inversed_idx % 4:
233 |                         if digit > 1:
234 |                             tmp += self.literal[digit]
235 |                         if digit:
236 |                             tmp += self.small_scale[inversed_idx % 4]
237 |                     else:
238 |                         if digit or length == 1:
239 |                             tmp += self.literal[digit]
240 |                         if tmp:
241 |                             tmp += self.large_scale[inversed_idx // 4]
242 |                         if length == 5 and digit == 1 and inversed_idx == 4:
243 |                             tmp = tmp[1:]
244 |                         korean += tmp
245 |                         tmp = str()
246 |             korean += tmp
247 |             return korean
248 | 
249 |         def individual_read(self, frag):
250 |             korean = str()
251 |             for digit in frag:
252 |                 if digit == '_':
253 |                     korean += '공'
254 |                 else:
255 |                     korean += self.literal[int(digit)]
256 |             return korean
257 | 
258 |         result = self.sentence
259 |         if preprocessed_data is None:
260 |             return None
261 | 
262 |         for each in preprocessed_data:
263 |             reading_method, target_seq, original = each
264 |             readed = str()
265 |             for idx, frag in enumerate(target_seq):
266 |                 if reading_method[idx] == 'literal':
267 |                     readed += literal_read(self, frag)
268 |                 if reading_method[idx] == 'individual':
269 |                     readed += individual_read(self, frag)
270 |                 if reading_method[idx] == 'point':
271 |                     readed += ' 점 '
272 |                 if reading_method[idx] == 'of':
273 |                     readed += '의 '
274 |                 if reading_method[idx] == 'super':
275 |                     readed += ' 승'
276 |             result = result.replace(original, readed, 1)
277 |         return result
278 | 
279 |     def convert(self, sentence):
280 |         return self._read(self._preprocess(self._detect(sentence)))


--------------------------------------------------------------------------------
/preprocess/special.csv:
--------------------------------------------------------------------------------
  1 | special,replace
  2 |  ,
  3 | .,.
  4 | “,
  5 | """",
  6 | -,
  7 | ",",
  8 | ·,.
  9 | ”,
 10 | ㎏,킬로그램
 11 | ㎞,킬로미터
 12 | ?,?
 13 | ‘,
 14 | ’,
 15 | ;,
 16 | %,퍼센트
 17 | !,!
 18 | ㎓,기가헤르츠
 19 | ',
 20 | ♪,
 21 | ∼, 
 22 | [,
 23 | ],
 24 | &,
 25 | ~, 
 26 | /,
 27 | <,
 28 | >,
 29 | 「,
 30 | 」,
 31 | $,달러
 32 | :,
 33 | _,
 34 | 《,
 35 | 》,
 36 | ㈜,
 37 | Ⅱ,
 38 | ㎡,제곱미터
 39 | ¾,사 분의 삼
 40 | ,
 41 | …,
 42 | ㎜,밀리미터
 43 | =,는
 44 | #,샾
 45 | 『,
 46 | 』,
 47 | ‧,
 48 | ♬,
 49 | ㏊,
 50 | @,앳
 51 | ℉,화씨
 52 | ㎥,세제곱미터
 53 | Ⅲ,
 54 | →,
 55 | 〈,
 56 | 〉,
 57 | ㎝,센티미터
 58 | {,
 59 | \,원
 60 | },
 61 | ¥,
 62 | ㎾,킬로와트
 63 | Ⅰ,
 64 | ․,.
 65 | ㈔,
 66 | ㎖,밀리리터
 67 | ○,
 68 | ㎍,마이크로그램
 69 | ​,
 70 | ㎿,메가와트
 71 | ℃,도
 72 | ㎳,밀리세컨드
 73 | +,플러스
 74 | ™,
 75 | ＜,
 76 | ＞,
 77 | *,
 78 | ㎢,제곱킬로미터
 79 | ⅓,삼 분의 일
 80 | ＆,
 81 | ^,
 82 | ㎚,나노미터
 83 | ‭,
 84 | √,
 85 | ㎎,밀리그램
 86 | 〮,
 87 | ♡,
 88 | ∙,
 89 | ㎛,마이크로미터
 90 | °,도
 91 | –,
 92 | ,
 93 | Ⅹ,
 94 | ″,
 95 | ±,플러스 마이너스
 96 | ㎘,킬로리터
 97 | ≪,
 98 | ≫,
 99 | £,
100 | ×,
101 | �,
102 | ㎉,킬로칼로리
103 | ★,
104 | □,
105 | ˚,
106 | ㎧,미터 퍼 세크
107 | ㎈,칼로리
108 | •,
109 | ′,
110 | ¡,
111 | 〓,
112 | 【,
113 | 】,
114 | 〔,
115 | 〕,
116 | ・,
117 | ˙,
118 | ㏈,데시벨
119 | Ⅵ,
120 | Ⅳ,
121 | ？,?
122 | ⅔,삼 분의 이
123 | 、,
124 | 〜,
125 | ♥,
126 | `,
127 | ®,
128 | 〇,
129 | ㎠,제곱센티미터
130 | ㏄,씨씨
131 | ▒,
132 | ‬,
133 | —,
134 | ｢,
135 | ｣,
136 | ！,!
137 | ―,
138 | ♩,
139 | ⓛ,
140 | ㎒,메가헤르츠
141 | ▶,
142 | ：,
143 | ˜,
144 | ´,
145 | ㎽,밀리와트
146 | ,
147 | ～,
148 | ┃,
149 | ·,
150 | ⼐,
151 | ㎃,밀리암페어
152 | ％,퍼센트
153 | ◆,
154 | －,
155 | ⇒,
156 | ‰,퍼밀
157 | （,
158 | ）,
159 | ㍱,헥토파스칼
160 | ━,
161 | ㎐,헤르츠
162 | ≡,
163 | ¿,
164 | 。,
165 | €,
166 | ¶,
167 | ･,
168 | ㎔,테라헤르츠
169 | ※,
170 | Ⅸ,
171 | ⑯,
172 | –,
173 | ，,
174 | ↗,
175 | ↘,
176 | ＊,
177 | ㎸,킬로볼트
178 | ㉦,
179 | ½,이 분의 일
180 | ⟪,
181 | ⟫,
182 | ƒ,
183 | …,
184 | Ⅶ,
185 | ．,
186 | §,
187 | ㎨,미터 퍼 세크 제곱
188 | △,
189 | ☆,
190 | ÷,나누기
191 | ∪,
192 | ㎩,파스칼
193 | ,
194 | 〃,
195 | ㎫,메가파스칼
196 | ￡,
197 | ㎗,
198 | ■,
199 | ,
200 | ,
201 | ‵,
202 | ⓧ,
203 | ­,
204 | ¢,
205 | ♫,
206 | €,
207 | ㎟,제곱밀리미터
208 | ➡,
209 | ⃗,
210 | ㎑,킬로헤르츠
211 | ​,
212 | 


--------------------------------------------------------------------------------
/preprocess/train_labels.csv:
--------------------------------------------------------------------------------
   1 | id,char,freq
   2 | 0, ,5774462
   3 | 1,.,640924
   4 | 2,그,556373
   5 | 3,이,509291
   6 | 4,는,374559
   7 | 5,아,370444
   8 | 6,가,369698
   9 | 7,고,356378
  10 | 8,어,333842
  11 | 9,거,306987
  12 | 10,지,276453
  13 | 11,데,249269
  14 | 12,?,235024
  15 | 13,나,229646
  16 | 14,하,226136
  17 | 15,다,221216
  18 | 16,서,211193
  19 | 17,에,204330
  20 | 18,도,190561
  21 | 19,게,177140
  22 | 20,니,173284
  23 | 21,",",152938
  24 | 22,기,149467
  25 | 23,은,144674
  26 | 24,면,142025
  27 | 25,야,137553
  28 | 26,있,133155
  29 | 27,한,121564
  30 | 28,을,121048
  31 | 29,까,119483
  32 | 30,해,115148
  33 | 31,리,111855
  34 | 32,라,111479
  35 | 33,래,105784
  36 | 34,사,100533
  37 | 35,근,99781
  38 | 36,들,99447
  39 | 37,안,97043
  40 | 38,로,91847
  41 | 39,일,88319
  42 | 40,뭐,87328
  43 | 41,내,85968
  44 | 42,보,82911
  45 | 43,제,80874
  46 | 44,같,79626
  47 | 45,자,76298
  48 | 46,만,76093
  49 | 47,시,72836
  50 | 48,런,70919
  51 | 49,너,69192
  52 | 50,대,68756
  53 | 51,때,67179
  54 | 52,되,66237
  55 | 53,으,66106
  56 | 54,진,62831
  57 | 55,를,61802
  58 | 56,잖,61455
  59 | 57,오,60782
  60 | 58,러,60629
  61 | 59,인,60234
  62 | 60,막,59994
  63 | 61,무,58705
  64 | 62,었,58385
  65 | 63,구,57294
  66 | 64,했,57209
  67 | 65,수,56787
  68 | 66,간,55275
  69 | 67,애,54476
  70 | 68,우,53539
  71 | 69,요,53234
  72 | 70,마,53125
  73 | 71,생,52815
  74 | 72,렇,50798
  75 | 73,냥,49989
  76 | 74,짜,49581
  77 | 75,주,48969
  78 | 76,없,48392
  79 | 77,말,47929
  80 | 78,학,46285
  81 | 79,스,46225
  82 | 80,더,44487
  83 | 81,많,43607
  84 | 82,원,41379
  85 | 83,음,41348
  86 | 84,정,39775
  87 | 85,겠,39691
  88 | 86,여,39203
  89 | 87,먹,39194
  90 | 88,금,38720
  91 | 89,든,38476
  92 | 90,부,38398
  93 | 91,할,38262
  94 | 92,전,36575
  95 | 93,번,36375
  96 | 94,좋,36363
  97 | 95,랑,36081
  98 | 96,네,35514
  99 | 97,람,33799
 100 | 98,약,33412
 101 | 99,건,33371
 102 | 100,각,32167
 103 | 101,좀,31738
 104 | 102,알,30893
 105 | 103,잘,30132
 106 | 104,걸,29634
 107 | 105,모,29629
 108 | 106,것,28482
 109 | 107,상,28247
 110 | 108,려,28218
 111 | 109,장,27856
 112 | 110,히,27705
 113 | 111,않,27305
 114 | 112,맞,27202
 115 | 113,던,27082
 116 | 114,르,26286
 117 | 115,교,26116
 118 | 116,바,25994
 119 | 117,냐,25742
 120 | 118,드,25702
 121 | 119,십,25654
 122 | 120,날,25556
 123 | 121,치,25287
 124 | 122,비,25278
 125 | 123,단,25129
 126 | 124,동,25047
 127 | 125,또,24720
 128 | 126,못,24528
 129 | 127,저,24074
 130 | 128,얘,23990
 131 | 129,중,23851
 132 | 130,의,23607
 133 | 131,난,23318
 134 | 132,엄,23057
 135 | 133,봤,22930
 136 | 134,걔,22732
 137 | 135,화,22593
 138 | 136,응,22254
 139 | 137,싶,21756
 140 | 138,갔,21628
 141 | 139,았,21052
 142 | 140,집,20850
 143 | 141,왜,20801
 144 | 142,계,20757
 145 | 143,공,20620
 146 | 144,긴,20547
 147 | 145,신,20371
 148 | 146,적,20244
 149 | 147,연,20225
 150 | 148,직,20061
 151 | 149,실,19467
 152 | 150,영,19454
 153 | 151,미,19444
 154 | 152,봐,18931
 155 | 153,분,18893
 156 | 154,테,18829
 157 | 155,년,18669
 158 | 156,트,18654
 159 | 157,문,18230
 160 | 158,와,18114
 161 | 159,돼,18028
 162 | 160,물,17889
 163 | 161,예,17864
 164 | 162,터,17722
 165 | 163,세,17719
 166 | 164,럼,17521
 167 | 165,청,17479
 168 | 166,차,17455
 169 | 167,친,17355
 170 | 168,개,17355
 171 | 169,삼,17242
 172 | 170,국,17224
 173 | 171,두,17129
 174 | 172,소,17125
 175 | 173,살,16893
 176 | 174,재,16635
 177 | 175,운,15949
 178 | 176,쫌,15780
 179 | 177,유,15516
 180 | 178,속,15255
 181 | 179,명,15190
 182 | 180,랬,15155
 183 | 181,본,15148
 184 | 182,갈,15084
 185 | 183,방,15069
 186 | 184,돈,14998
 187 | 185,타,14919
 188 | 186,처,14908
 189 | 187,빠,14851
 190 | 188,과,14843
 191 | 189,식,14743
 192 | 190,디,14633
 193 | 191,배,14403
 194 | 192,피,14093
 195 | 193,뭔,14093
 196 | 194,선,13929
 197 | 195,남,13909
 198 | 196,경,13868
 199 | 197,달,13787
 200 | 198,언,13579
 201 | 199,받,13519
 202 | 200,심,13409
 203 | 201,월,13367
 204 | 202,버,13339
 205 | 203,왔,13265
 206 | 204,느,13223
 207 | 205,점,12960
 208 | 206,올,12923
 209 | 207,업,12861
 210 | 208,른,12801
 211 | 209,성,12717
 212 | 210,회,12591
 213 | 211,조,12570
 214 | 212,워,12424
 215 | 213,따,12410
 216 | 214,행,12350
 217 | 215,반,12158
 218 | 216,님,11998
 219 | 217,딱,11908
 220 | 218,관,11828
 221 | 219,입,11275
 222 | 220,카,11235
 223 | 221,당,11068
 224 | 222,였,10977
 225 | 223,케,10576
 226 | 224,쪽,10430
 227 | 225,천,10384
 228 | 226,작,10381
 229 | 227,누,10336
 230 | 228,열,10252
 231 | 229,얼,10250
 232 | 230,울,10246
 233 | 231,찮,10231
 234 | 232,된,10191
 235 | 233,별,10159
 236 | 234,떻,10108
 237 | 235,머,9876
 238 | 236,쓰,9853
 239 | 237,위,9841
 240 | 238,크,9838
 241 | 239,노,9799
 242 | 240,괜,9735
 243 | 241,강,9698
 244 | 242,편,9668
 245 | 243,몰,9623
 246 | 244,맛,9382
 247 | 245,준,9342
 248 | 246,줄,9294
 249 | 247,파,9282
 250 | 248,백,9252
 251 | 249,매,9181
 252 | 250,산,9160
 253 | 251,술,9142
 254 | 252,힘,9056
 255 | 253,프,9019
 256 | 254,즘,8997
 257 | 255,임,8969
 258 | 256,체,8888
 259 | 257,형,8790
 260 | 258,몇,8742
 261 | 259,맨,8712
 262 | 260,새,8623
 263 | 261,용,8571
 264 | 262,키,8547
 265 | 263,통,8410
 266 | 264,양,8363
 267 | 265,끝,8361
 268 | 266,싸,8328
 269 | 267,볼,8188
 270 | 268,혼,8179
 271 | 269,온,8132
 272 | 270,등,8123
 273 | 271,길,8067
 274 | 272,될,8033
 275 | 273,밌,7998
 276 | 274,육,7924
 277 | 275,늘,7909
 278 | 276,슨,7835
 279 | 277,됐,7738
 280 | 278,놀,7707
 281 | 279,외,7608
 282 | 280,팔,7601
 283 | 281,져,7551
 284 | 282,레,7485
 285 | 283,억,7461
 286 | 284,발,7450
 287 | 285,결,7412
 288 | 286,초,7290
 289 | 287,감,7180
 290 | 288,군,7174
 291 | 289,호,7173
 292 | 290,름,7146
 293 | 291,솔,7079
 294 | 292,닌,7051
 295 | 293,밖,7013
 296 | 294,불,7007
 297 | 295,밥,6784
 298 | 296,포,6676
 299 | 297,싫,6631
 300 | 298,완,6582
 301 | 299,갖,6511
 302 | 300,겨,6468
 303 | 301,질,6453
 304 | 302,토,6448
 305 | 303,험,6417
 306 | 304,색,6371
 307 | 305,떤,6352
 308 | 306,역,6340
 309 | 307,티,6319
 310 | 308,갑,6316
 311 | 309,목,6262
 312 | 310,린,6256
 313 | 311,추,6204
 314 | 312,격,6174
 315 | 313,후,6119
 316 | 314,확,6095
 317 | 315,루,6079
 318 | 316,민,6024
 319 | 317,끼,6023
 320 | 318,칠,6019
 321 | 319,돌,5997
 322 | 320,찍,5956
 323 | 321,쪼,5946
 324 | 322,깐,5788
 325 | 323,필,5786
 326 | 324,빨,5693
 327 | 325,났,5657
 328 | 326,락,5561
 329 | 327,박,5537
 330 | 328,끔,5483
 331 | 329,낌,5403
 332 | 330,럴,5397
 333 | 331,취,5344
 334 | 332,복,5315
 335 | 333,둘,5264
 336 | 334,페,5217
 337 | 335,렸,5198
 338 | 336,써,5197
 339 | 337,줘,5173
 340 | 338,급,5067
 341 | 339,력,5065
 342 | 340,잡,5030
 343 | 341,씩,5006
 344 | 342,찾,4990
 345 | 343,놓,4987
 346 | 344,최,4894
 347 | 345,코,4891
 348 | 346,넘,4870
 349 | 347,졌,4803
 350 | 348,섯,4799
 351 | 349,브,4793
 352 | 350,현,4764
 353 | 351,눈,4760
 354 | 352,항,4751
 355 | 353,귀,4708
 356 | 354,설,4688
 357 | 355,벌,4666
 358 | 356,담,4647
 359 | 357,앞,4640
 360 | 358,책,4630
 361 | 359,절,4629
 362 | 360,플,4523
 363 | 361,폰,4513
 364 | 362,태,4496
 365 | 363,종,4487
 366 | 364,옛,4450
 367 | 365,증,4413
 368 | 366,튼,4411
 369 | 367,글,4408
 370 | 368,습,4383
 371 | 369,병,4377
 372 | 370,론,4373
 373 | 371,출,4364
 374 | 372,능,4354
 375 | 373,침,4345
 376 | 374,순,4339
 377 | 375,줬,4308
 378 | 376,평,4303
 379 | 377,메,4287
 380 | 378,똑,4281
 381 | 379,커,4261
 382 | 380,엔,4248
 383 | 381,꾸,4230
 384 | 382,란,4194
 385 | 383,듣,4083
 386 | 384,씨,4009
 387 | 385,큰,4002
 388 | 386,표,3995
 389 | 387,잠,3942
 390 | 388,먼,3942
 391 | 389,쁘,3840
 392 | 390,활,3820
 393 | 391,합,3787
 394 | 392,접,3732
 395 | 393,럽,3722
 396 | 394,옷,3705
 397 | 395,쳐,3690
 398 | 396,손,3689
 399 | 397,붙,3645
 400 | 398,망,3640
 401 | 399,죽,3609
 402 | 400,투,3606
 403 | 401,족,3603
 404 | 402,셨,3589
 405 | 403,참,3572
 406 | 404,떨,3567
 407 | 405,웃,3533
 408 | 406,졸,3516
 409 | 407,쉬,3492
 410 | 408,뭘,3447
 411 | 409,변,3406
 412 | 410,릴,3374
 413 | 411,웠,3293
 414 | 412,홍,3267
 415 | 413,즈,3265
 416 | 414,랐,3245
 417 | 415,독,3243
 418 | 416,충,3239
 419 | 417,짝,3217
 420 | 418,떡,3197
 421 | 419,뒤,3195
 422 | 420,휴,3161
 423 | 421,셔,3142
 424 | 422,넣,3135
 425 | 423,쨌,3075
 426 | 424,악,3073
 427 | 425,패,3049
 428 | 426,빼,3041
 429 | 427,슬,2983
 430 | 428,특,2975
 431 | 429,꺼,2970
 432 | 430,숙,2951
 433 | 431,쯤,2934
 434 | 432,텐,2905
 435 | 433,창,2901
 436 | 434,겼,2888
 437 | 435,굴,2869
 438 | 436,판,2863
 439 | 437,죠,2851
 440 | 438,답,2820
 441 | 439,희,2816
 442 | 440,허,2815
 443 | 441,옆,2798
 444 | 442,료,2791
 445 | 443,닐,2790
 446 | 444,택,2769
 447 | 445,림,2760
 448 | 446,읽,2742
 449 | 447,핸,2733
 450 | 448,축,2730
 451 | 449,풀,2716
 452 | 450,틀,2712
 453 | 451,몸,2694
 454 | 452,골,2690
 455 | 453,황,2635
 456 | 454,켜,2635
 457 | 455,익,2625
 458 | 456,베,2613
 459 | 457,북,2600
 460 | 458,법,2578
 461 | 459,늦,2578
 462 | 460,함,2568
 463 | 461,랜,2555
 464 | 462,꼬,2555
 465 | 463,향,2547
 466 | 464,석,2541
 467 | 465,환,2533
 468 | 466,슷,2529
 469 | 467,품,2518
 470 | 468,혀,2513
 471 | 469,블,2512
 472 | 470,쓸,2503
 473 | 471,채,2472
 474 | 472,며,2470
 475 | 473,욕,2463
 476 | 474,권,2450
 477 | 475,검,2445
 478 | 476,굳,2428
 479 | 477,록,2425
 480 | 478,톡,2408
 481 | 479,김,2408
 482 | 480,넌,2383
 483 | 481,깨,2375
 484 | 482,션,2374
 485 | 483,캐,2369
 486 | 484,송,2339
 487 | 485,녀,2336
 488 | 486,탈,2327
 489 | 487,광,2321
 490 | 488,혹,2313
 491 | 489,퍼,2300
 492 | 490,뽑,2278
 493 | 491,철,2265
 494 | 492,째,2249
 495 | 493,움,2232
 496 | 494,밤,2231
 497 | 495,꼭,2226
 498 | 496,샀,2224
 499 | 497,끊,2212
 500 | 498,땐,2203
 501 | 499,깔,2179
 502 | 500,멀,2145
 503 | 501,높,2141
 504 | 502,께,2140
 505 | 503,큼,2121
 506 | 504,녁,2104
 507 | 505,곳,2082
 508 | 506,잔,2070
 509 | 507,쉽,2070
 510 | 508,짐,2067
 511 | 509,암,2063
 512 | 510,극,2061
 513 | 511,련,2056
 514 | 512,떠,2056
 515 | 513,벽,2049
 516 | 514,헤,2047
 517 | 515,C,2040
 518 | 516,끄,2024
 519 | 517,곱,2015
 520 | 518,승,2011
 521 | 519,봉,2009
 522 | 520,착,2006
 523 | 521,촌,1990
 524 | 522,껴,1986
 525 | 523,딩,1983
 526 | 524,류,1978
 527 | 525,뜨,1970
 528 | 526,넷,1941
 529 | 527,놨,1922
 530 | 528,궁,1894
 531 | 529,논,1882
 532 | 530,곤,1875
 533 | 531,클,1869
 534 | 532,싼,1859
 535 | 533,앉,1854
 536 | 534,컴,1849
 537 | 535,맥,1841
 538 | 536,팀,1830
 539 | 537,썼,1818
 540 | 538,낫,1801
 541 | 539,튜,1788
 542 | 540,걱,1786
 543 | 541,쁜,1770
 544 | 542,킨,1760
 545 | 543,빌,1752
 546 | 544,쿠,1748
 547 | 545,찌,1738
 548 | 546,쌤,1719
 549 | 547,T,1715
 550 | 548,밀,1711
 551 | 549,빵,1702
 552 | 550,냈,1702
 553 | 551,센,1691
 554 | 552,딴,1688
 555 | 553,쩌,1678
 556 | 554,딸,1678
 557 | 555,걍,1596
 558 | 556,획,1588
 559 | 557,씬,1582
 560 | 558,챙,1541
 561 | 559,첫,1536
 562 | 560,범,1530
 563 | 561,핑,1519
 564 | 562,굉,1519
 565 | 563,쩔,1514
 566 | 564,팅,1507
 567 | 565,긍,1486
 568 | 566,탄,1471
 569 | 567,덟,1470
 570 | 568,퇴,1469
 571 | 569,뛰,1469
 572 | 570,층,1467
 573 | 571,춰,1454
 574 | 572,훨,1447
 575 | 573,찬,1439
 576 | 574,듯,1424
 577 | 575,S,1396
 578 | 576,왕,1392
 579 | 577,텔,1385
 580 | 578,뉴,1382
 581 | 579,렌,1377
 582 | 580,탕,1374
 583 | 581,짓,1371
 584 | 582,밑,1365
 585 | 583,헬,1358
 586 | 584,존,1339
 587 | 585,립,1323
 588 | 586,녔,1318
 589 | 587,꼈,1305
 590 | 588,빡,1304
 591 | 589,낮,1287
 592 | 590,견,1282
 593 | 591,링,1281
 594 | 592,볶,1271
 595 | 593,낙,1271
 596 | 594,릭,1267
 597 | 595,젠,1263
 598 | 596,퓨,1262
 599 | 597,츠,1256
 600 | 598,맘,1252
 601 | 599,놔,1247
 602 | 600,렵,1241
 603 | 601,땜,1232
 604 | 602,쇼,1224
 605 | 603,값,1215
 606 | 604,닭,1203
 607 | 605,깝,1200
 608 | 606,픈,1194
 609 | 607,탁,1183
 610 | 608,쓴,1179
 611 | 609,농,1172
 612 | 610,량,1166
 613 | 611,염,1156
 614 | 612,홉,1144
 615 | 613,척,1130
 616 | 614,겁,1129
 617 | 615,콘,1127
 618 | 616,섭,1125
 619 | 617,냄,1125
 620 | 618,P,1125
 621 | 619,효,1124
 622 | 620,규,1124
 623 | 621,꿈,1121
 624 | 622,곡,1093
 625 | 623,액,1090
 626 | 624,쎄,1077
 627 | 625,덜,1075
 628 | 626,턴,1065
 629 | 627,킹,1061
 630 | 628,훈,1057
 631 | 629,쳤,1054
 632 | 630,널,1047
 633 | 631,멋,1037
 634 | 632,꿀,1034
 635 | 633,깜,1019
 636 | 634,짧,1014
 637 | 635,롤,1013
 638 | 636,낼,1012
 639 | 637,꽤,1003
 640 | 638,총,984
 641 | 639,램,984
 642 | 640,덕,980
 643 | 641,믄,974
 644 | 642,믿,972
 645 | 643,흥,970
 646 | 644,롱,967
 647 | 645,뜻,962
 648 | 646,짤,958
 649 | 647,쌍,957
 650 | 648,컨,953
 651 | 649,셋,952
 652 | 650,잤,950
 653 | 651,닥,950
 654 | 652,웬,946
 655 | 653,엽,944
 656 | 654,혜,939
 657 | 655,찰,935
 658 | 656,뻐,935
 659 | 657,뿌,934
 660 | 658,빈,934
 661 | 659,꿔,934
 662 | 660,낸,932
 663 | 661,뻔,928
 664 | 662,쌓,926
 665 | 663,즐,919
 666 | 664,튀,914
 667 | 665,겹,909
 668 | 666,득,899
 669 | 667,끌,896
 670 | 668,M,880
 671 | 669,V,877
 672 | 670,녹,876
 673 | 671,푸,870
 674 | 672,쭉,863
 675 | 673,싱,858
 676 | 674,팬,857
 677 | 675,A,854
 678 | 676,!,841
 679 | 677,념,836
 680 | 678,맡,825
 681 | 679,쟁,814
 682 | 680,엑,810
 683 | 681,켓,809
 684 | 682,뀌,808
 685 | 683,털,803
 686 | 684,풍,802
 687 | 685,웨,799
 688 | 686,땡,792
 689 | 687,롯,791
 690 | 688,롭,788
 691 | 689,젊,781
 692 | 690,넓,778
 693 | 691,멘,777
 694 | 692,냉,772
 695 | 693,칼,771
 696 | 694,잉,768
 697 | 695,빙,768
 698 | 696,뿐,767
 699 | 697,옮,761
 700 | 698,젤,760
 701 | 699,B,757
 702 | 700,죄,756
 703 | 701,탔,752
 704 | 702,샤,746
 705 | 703,홀,745
 706 | 704,떼,743
 707 | 705,줌,738
 708 | 706,징,734
 709 | 707,폭,727
 710 | 708,G,721
 711 | 709,킬,713
 712 | 710,흔,712
 713 | 711,딜,711
 714 | 712,슈,703
 715 | 713,율,700
 716 | 714,즌,697
 717 | 715,씀,694
 718 | 716,앙,689
 719 | 717,눠,688
 720 | 718,콩,686
 721 | 719,얻,684
 722 | 720,숨,682
 723 | 721,닝,673
 724 | 722,꽃,668
 725 | 723,쌀,667
 726 | 724,컬,666
 727 | 725,춤,666
 728 | 726,c,664
 729 | 727,뚫,663
 730 | 728,엠,661
 731 | 729,몬,659
 732 | 730,D,658
 733 | 731,흐,656
 734 | 732,앤,646
 735 | 733,똥,645
 736 | 734,콜,644
 737 | 735,델,635
 738 | 736,렀,628
 739 | 737,폐,627
 740 | 738,엘,624
 741 | 739,쁠,623
 742 | 740,랄,622
 743 | 741,걘,621
 744 | 742,벤,619
 745 | 743,봄,611
 746 | 744,왠,609
 747 | 745,씻,609
 748 | 746,률,608
 749 | 747,켰,600
 750 | 748,짱,600
 751 | 749,웹,599
 752 | 750,압,599
 753 | 751,럭,596
 754 | 752,땅,596
 755 | 753,멍,595
 756 | 754,랩,595
 757 | 755,댓,595
 758 | 756,깊,595
 759 | 757,뮤,592
 760 | 758,령,590
 761 | 759,릿,589
 762 | 760,낀,589
 763 | 761,윤,586
 764 | 762,옥,584
 765 | 763,룸,582
 766 | 764,딘,579
 767 | 765,객,578
 768 | 766,댄,576
 769 | 767,컵,574
 770 | 768,폴,573
 771 | 769,쟤,570
 772 | 770,뷰,569
 773 | 771,템,568
 774 | 772,덴,567
 775 | 773,눌,559
 776 | 774,캠,558
 777 | 775,홈,557
 778 | 776,삶,557
 779 | 777,삭,555
 780 | 778,벨,555
 781 | 779,엉,552
 782 | 780,헐,549
 783 | 781,벅,545
 784 | 782,벗,544
 785 | 783,혈,543
 786 | 784,밍,539
 787 | 785,셀,536
 788 | 786,낭,534
 789 | 787,춥,533
 790 | 788,릉,533
 791 | 789,t,533
 792 | 790,잃,532
 793 | 791,I,529
 794 | 792,놈,526
 795 | 793,춘,524
 796 | 794,찜,520
 797 | 795,R,519
 798 | 796,걷,518
 799 | 797,삐,515
 800 | 798,헌,510
 801 | 799,딨,510
 802 | 800,빛,504
 803 | 801,흘,503
 804 | 802,닫,502
 805 | 803,균,502
 806 | 804,p,495
 807 | 805,L,494
 808 | 806,좌,492
 809 | 807,껄,491
 810 | 808,펜,489
 811 | 809,N,487
 812 | 810,싹,486
 813 | 811,탑,485
 814 | 812,쏘,483
 815 | 813,O,482
 816 | 814,픽,480
 817 | 815,덩,477
 818 | 816,햄,476
 819 | 817,큐,473
 820 | 818,힐,472
 821 | 819,곧,471
 822 | 820,낳,470
 823 | 821,힌,468
 824 | 822,팩,468
 825 | 823,뒷,468
 826 | 824,툰,467
 827 | 825,섬,465
 828 | 826,꽂,463
 829 | 827,례,462
 830 | 828,핫,460
 831 | 829,섞,460
 832 | 830,촬,458
 833 | 831,흰,457
 834 | 832,둥,455
 835 | 833,K,450
 836 | 834,괴,449
 837 | 835,s,448
 838 | 836,핀,446
 839 | 837,꿨,444
 840 | 838,틱,441
 841 | 839,밝,441
 842 | 840,랙,440
 843 | 841,땠,440
 844 | 842,둔,440
 845 | 843,슴,439
 846 | 844,첨,438
 847 | 845,밴,432
 848 | 846,렁,431
 849 | 847,칭,429
 850 | 848,묻,428
 851 | 849,뜬,425
 852 | 850,깎,424
 853 | 851,엇,423
 854 | 852,컸,421
 855 | 853,퀴,420
 856 | 854,납,418
 857 | 855,협,417
 858 | 856,몽,416
 859 | 857,꼐,415
 860 | 858,떴,414
 861 | 859,썰,410
 862 | 860,찐,407
 863 | 861,꼴,407
 864 | 862,갠,406
 865 | 863,턱,405
 866 | 864,틴,398
 867 | 865,낄,398
 868 | 866,뒀,397
 869 | 867,끗,396
 870 | 868,꼼,395
 871 | 869,F,395
 872 | 870,샵,394
 873 | 871,휘,392
 874 | 872,뼈,390
 875 | 873,뚜,389
 876 | 874,쩍,388
 877 | 875,팡,386
 878 | 876,멜,386
 879 | 877,톤,385
 880 | 878,앨,385
 881 | 879,탐,384
 882 | 880,칸,384
 883 | 881,끓,383
 884 | 882,뚱,381
 885 | 883,닮,378
 886 | 884,깃,375
 887 | 885,짬,374
 888 | 886,빤,371
 889 | 887,측,370
 890 | 888,혔,369
 891 | 889,꽁,369
 892 | 890,펴,368
 893 | 891,앴,368
 894 | 892,겸,368
 895 | 893,쿨,367
 896 | 894,릇,363
 897 | 895,얀,362
 898 | 896,쿄,358
 899 | 897,컷,358
 900 | 898,팠,356
 901 | 899,끈,355
 902 | 900,렴,354
 903 | 901,잊,352
 904 | 902,덤,350
 905 | 903,갤,342
 906 | 904,븐,340
 907 | 905,흡,337
 908 | 906,덮,337
 909 | 907,씹,335
 910 | 908,뽀,335
 911 | 909,뚝,335
 912 | 910,갚,335
 913 | 911,찔,334
 914 | 912,댔,333
 915 | 913,혁,332
 916 | 914,띠,328
 917 | 915,벼,327
 918 | 916,얇,324
 919 | 917,뺐,324
 920 | 918,팝,323
 921 | 919,잇,322
 922 | 920,왼,322
 923 | 921,낚,321
 924 | 922,칙,316
 925 | 923,겉,316
 926 | 924,뜯,313
 927 | 925,닦,312
 928 | 926,짠,311
 929 | 927,썹,310
 930 | 928,뷔,310
 931 | 929,묶,310
 932 | 930,꾼,306
 933 | 931,빅,305
 934 | 932,땄,305
 935 | 933,캡,304
 936 | 934,묘,304
 937 | 935,샘,303
 938 | 936,묵,303
 939 | 937,a,302
 940 | 938,쭈,300
 941 | 939,b,300
 942 | 940,겪,299
 943 | 941,둬,298
 944 | 942,J,298
 945 | 943,쫄,296
 946 | 944,랫,296
 947 | 945,뀐,296
 948 | 946,흑,295
 949 | 947,댕,295
 950 | 948,꽉,295
 951 | 949,곰,294
 952 | 950,붕,293
 953 | 951,땀,292
 954 | 952,릎,290
 955 | 953,뽕,289
 956 | 954,쥐,288
 957 | 955,렉,287
 958 | 956,숭,283
 959 | 957,샐,283
 960 | 958,v,282
 961 | 959,렛,281
 962 | 960,녕,281
 963 | 961,힙,280
 964 | 962,쫙,279
 965 | 963,촉,278
 966 | 964,쩜,277
 967 | 965,긋,277
 968 | 966,샌,276
 969 | 967,o,275
 970 | 968,쫓,273
 971 | 969,쩐,273
 972 | 970,헷,272
 973 | 971,X,268
 974 | 972,웅,267
 975 | 973,뺏,267
 976 | 974,쵸,266
 977 | 975,쪘,266
 978 | 976,랍,266
 979 | 977,E,266
 980 | 978,좁,265
 981 | 979,앱,265
 982 | 980,썸,264
 983 | 981,냅,264
 984 | 982,펙,263
 985 | 983,늙,263
 986 | 984,껌,261
 987 | 985,n,261
 988 | 986,e,261
 989 | 987,랭,260
 990 | 988,귤,260
 991 | 989,찢,259
 992 | 990,닿,259
 993 | 991,띄,258
 994 | 992,긁,255
 995 | 993,귄,253
 996 | 994,굽,253
 997 | 995,갓,253
 998 | 996,캔,252
 999 | 997,멈,252
1000 | 998,욱,250
1001 | 999,뺄,250
1002 | 1000,뇌,250
1003 | 1001,팟,249
1004 | 1002,쌌,248
1005 | 1003,룹,248
1006 | 1004,덥,248
1007 | 1005,폼,246
1008 | 1006,톱,244
1009 | 1007,듬,244
1010 | 1008,껍,244
1011 | 1009,흠,243
1012 | 1010,팍,243
1013 | 1011,맹,243
1014 | 1012,쉴,242
1015 | 1013,썩,240
1016 | 1014,밟,240
1017 | 1015,맵,237
1018 | 1016,돋,236
1019 | 1017,콤,235
1020 | 1018,맙,234
1021 | 1019,뱅,233
1022 | 1020,쫍,231
1023 | 1021,윗,229
1024 | 1022,뜩,229
1025 | 1023,찝,228
1026 | 1024,뺀,227
1027 | 1025,닷,226
1028 | 1026,넨,226
1029 | 1027,쌈,225
1030 | 1028,쩨,224
1031 | 1029,붓,224
1032 | 1030,쩡,223
1033 | 1031,믹,223
1034 | 1032,잼,221
1035 | 1033,r,221
1036 | 1034,쭐,220
1037 | 1035,엊,219
1038 | 1036,g,219
1039 | 1037,췄,217
1040 | 1038,룩,217
1041 | 1039,텀,215
1042 | 1040,쇠,213
1043 | 1041,숫,212
1044 | 1042,풋,210
1045 | 1043,쌩,208
1046 | 1044,쾌,207
1047 | 1045,볍,207
1048 | 1046,뤄,207
1049 | 1047,겐,207
1050 | 1048,m,207
1051 | 1049,펌,206
1052 | 1050,쪄,206
1053 | 1051,뻥,206
1054 | 1052,i,206
1055 | 1053,뻤,205
1056 | 1054,k,204
1057 | 1055,핵,203
1058 | 1056,셉,200
1059 | 1057,듀,198
1060 | 1058,닉,198
1061 | 1059,략,197
1062 | 1060,넉,197
1063 | 1061,딪,196
1064 | 1062,낯,195
1065 | 1063,텍,194
1066 | 1064,뱃,194
1067 | 1065,멤,194
1068 | 1066,윈,192
1069 | 1067,엎,192
1070 | 1068,뭉,192
1071 | 1069,젝,191
1072 | 1070,셜,190
1073 | 1071,빴,190
1074 | 1072,룰,190
1075 | 1073,앗,189
1076 | 1074,궈,189
1077 | 1075,윙,188
1078 | 1076,엥,188
1079 | 1077,d,185
1080 | 1078,꼽,183
1081 | 1079,챔,182
1082 | 1080,쉐,182
1083 | 1081,봇,182
1084 | 1082,푼,180
1085 | 1083,댁,179
1086 | 1084,칵,178
1087 | 1085,뿔,177
1088 | 1086,뺑,176
1089 | 1087,탱,175
1090 | 1088,쿼,174
1091 | 1089,갱,174
1092 | 1090,퉁,173
1093 | 1091,빔,173
1094 | 1092,썬,172
1095 | 1093,빽,172
1096 | 1094,둑,172
1097 | 1095,헛,169
1098 | 1096,빗,168
1099 | 1097,탓,167
1100 | 1098,륵,165
1101 | 1099,꼰,165
1102 | 1100,쎈,163
1103 | 1101,쥬,162
1104 | 1102,깡,162
1105 | 1103,퀄,161
1106 | 1104,빚,161
1107 | 1105,즉,160
1108 | 1106,삿,160
1109 | 1107,밭,160
1110 | 1108,u,160
1111 | 1109,혐,159
1112 | 1110,햇,159
1113 | 1111,툭,159
1114 | 1112,탠,158
1115 | 1113,샷,158
1116 | 1114,맣,157
1117 | 1115,껏,157
1118 | 1116,핏,156
1119 | 1117,앵,155
1120 | 1118,뜰,155
1121 | 1119,굿,155
1122 | 1120,U,155
1123 | 1121,섹,154
1124 | 1122,펑,153
1125 | 1123,맻,153
1126 | 1124,뀔,153
1127 | 1125,깥,152
1128 | 1126,뱀,150
1129 | 1127,뢰,150
1130 | 1128,껀,150
1131 | 1129,뉘,149
1132 | 1130,흉,146
1133 | 1131,틈,146
1134 | 1132,쏟,146
1135 | 1133,훔,143
1136 | 1134,쇄,142
1137 | 1135,뎅,142
1138 | 1136,칩,141
1139 | 1137,띵,140
1140 | 1138,푹,139
1141 | 1139,넥,139
1142 | 1140,퀘,137
1143 | 1141,훅,136
1144 | 1142,융,136
1145 | 1143,멸,136
1146 | 1144,냠,136
1147 | 1145,횟,134
1148 | 1146,찼,134
1149 | 1147,Y,134
1150 | 1148,룬,133
1151 | 1149,귈,133
1152 | 1150,H,133
1153 | 1151,젓,132
1154 | 1152,쏠,132
1155 | 1153,숲,132
1156 | 1154,냬,132
1157 | 1155,l,132
1158 | 1156,짰,130
1159 | 1157,멕,130
1160 | 1158,뇨,130
1161 | 1159,팽,128
1162 | 1160,깼,127
1163 | 1161,숏,126
1164 | 1162,굔,125
1165 | 1163,슐,123
1166 | 1164,쉰,123
1167 | 1165,얄,122
1168 | 1166,뱉,122
1169 | 1167,렬,122
1170 | 1168,굶,122
1171 | 1169,팁,121
1172 | 1170,츄,121
1173 | 1171,뭣,121
1174 | 1172,륙,121
1175 | 1173,횡,120
1176 | 1174,옹,119
1177 | 1175,뻘,117
1178 | 1176,옵,116
1179 | 1177,옴,116
1180 | 1178,얹,116
1181 | 1179,쑥,116
1182 | 1180,깄,116
1183 | 1181,므,115
1184 | 1182,찡,112
1185 | 1183,젖,112
1186 | 1184,꽈,112
1187 | 1185,틸,111
1188 | 1186,콕,111
1189 | 1187,첩,111
1190 | 1188,똘,111
1191 | 1189,쿵,110
1192 | 1190,왤,110
1193 | 1191,괌,110
1194 | 1192,밸,109
1195 | 1193,녜,109
1196 | 1194,갸,109
1197 | 1195,펀,108
1198 | 1196,칫,108
1199 | 1197,맺,108
1200 | 1198,탭,107
1201 | 1199,쁨,106
1202 | 1200,폈,105
1203 | 1201,펼,105
1204 | 1202,첼,105
1205 | 1203,숱,105
1206 | 1204,섰,105
1207 | 1205,킥,104
1208 | 1206,맑,103
1209 | 1207,랗,103
1210 | 1208,펐,102
1211 | 1209,넛,102
1212 | 1210,솜,101
1213 | 1211,벙,101
1214 | 1212,껑,101
1215 | 1213,f,101
1216 | 1214,룡,100
1217 | 1215,훌,99
1218 | 1216,x,98
1219 | 1217,쓱,97
1220 | 1218,늬,97
1221 | 1219,곽,97
1222 | 1220,y,97
1223 | 1221,욘,96
1224 | 1222,돔,96
1225 | 1223,겄,96
1226 | 1224,텝,95
1227 | 1225,훠,94
1228 | 1226,텅,94
1229 | 1227,씌,94
1230 | 1228,꺾,94
1231 | 1229,벚,93
1232 | 1230,렷,93
1233 | 1231,귓,93
1234 | 1232,찹,92
1235 | 1233,툴,91
1236 | 1234,깅,91
1237 | 1235,쭤,90
1238 | 1236,욜,90
1239 | 1237,얌,90
1240 | 1238,짖,89
1241 | 1239,옳,89
1242 | 1240,벳,88
1243 | 1241,뛸,88
1244 | 1242,깠,88
1245 | 1243,퍽,87
1246 | 1244,퀸,87
1247 | 1245,엮,87
1248 | 1246,삽,87
1249 | 1247,겟,87
1250 | 1248,왓,86
1251 | 1249,댈,86
1252 | 1250,샴,85
1253 | 1251,뻗,85
1254 | 1252,됨,84
1255 | 1253,얜,83
1256 | 1254,굵,83
1257 | 1255,눕,82
1258 | 1256,갇,82
1259 | 1257,셰,81
1260 | 1258,늫,81
1261 | 1259,텨,80
1262 | 1260,숍,80
1263 | 1261,뻑,80
1264 | 1262,됩,80
1265 | 1263,잎,79
1266 | 1264,뭇,79
1267 | 1265,퐁,78
1268 | 1266,팸,78
1269 | 1267,쯔,78
1270 | 1268,넜,78
1271 | 1269,깍,78
1272 | 1270,쌔,77
1273 | 1271,셈,77
1274 | 1272,읍,76
1275 | 1273,픔,75
1276 | 1274,펫,75
1277 | 1275,콧,75
1278 | 1276,얗,75
1279 | 1277,눅,75
1280 | 1278,j,74
1281 | 1279,쬐,73
1282 | 1280,렙,73
1283 | 1281,닙,73
1284 | 1282,슥,72
1285 | 1283,흙,71
1286 | 1284,쭝,71
1287 | 1285,짭,71
1288 | 1286,샹,71
1289 | 1287,릏,71
1290 | 1288,럿,71
1291 | 1289,덧,71
1292 | 1290,즙,70
1293 | 1291,늑,70
1294 | 1292,괄,70
1295 | 1293,킷,68
1296 | 1294,쿡,68
1297 | 1295,캉,68
1298 | 1296,둡,68
1299 | 1297,톨,67
1300 | 1298,엣,67
1301 | 1299,숟,67
1302 | 1300,낑,67
1303 | 1301,펭,66
1304 | 1302,왁,66
1305 | 1303,쏴,66
1306 | 1304,쏙,66
1307 | 1305,봅,66
1308 | 1306,멧,66
1309 | 1307,줏,65
1310 | 1308,뵈,65
1311 | 1309,쫑,64
1312 | 1310,륨,64
1313 | 1311,h,64
1314 | 1312,펄,63
1315 | 1313,짼,63
1316 | 1314,짚,63
1317 | 1315,껐,63
1318 | 1316,겜,63
1319 | 1317,싯,62
1320 | 1318,붐,62
1321 | 1319,렐,62
1322 | 1320,돗,62
1323 | 1321,팥,61
1324 | 1322,웰,61
1325 | 1323,륜,61
1326 | 1324,잣,60
1327 | 1325,슝,60
1328 | 1326,붉,60
1329 | 1327,윽,59
1330 | 1328,삘,59
1331 | 1329,딲,59
1332 | 1330,갯,58
1333 | 1331,횐,57
1334 | 1332,헨,57
1335 | 1333,캘,57
1336 | 1334,쩰,57
1337 | 1335,뤘,57
1338 | 1336,랴,57
1339 | 1337,껜,57
1340 | 1338,펠,56
1341 | 1339,킵,56
1342 | 1340,컹,56
1343 | 1341,렘,56
1344 | 1342,뛴,56
1345 | 1343,헝,55
1346 | 1344,씽,55
1347 | 1345,뮬,55
1348 | 1346,젯,54
1349 | 1347,샜,54
1350 | 1348,뿜,54
1351 | 1349,뒹,54
1352 | 1350,뎌,54
1353 | 1351,깬,54
1354 | 1352,챠,53
1355 | 1353,왈,53
1356 | 1354,뾰,53
1357 | 1355,뚤,53
1358 | 1356,꾹,53
1359 | 1357,갛,53
1360 | 1358,잌,52
1361 | 1359,엿,52
1362 | 1360,솥,52
1363 | 1361,벡,52
1364 | 1362,룻,52
1365 | 1363,꿍,52
1366 | 1364,곈,52
1367 | 1365,팜,51
1368 | 1366,튕,51
1369 | 1367,컥,51
1370 | 1368,첸,51
1371 | 1369,줍,51
1372 | 1370,섀,51
1373 | 1371,몫,51
1374 | 1372,뜸,51
1375 | 1373,깁,51
1376 | 1374,핬,50
1377 | 1375,쭘,50
1378 | 1376,쌰,50
1379 | 1377,넬,50
1380 | 1378,큘,49
1381 | 1379,쾅,49
1382 | 1380,캄,48
1383 | 1381,괘,48
1384 | 1382,쟀,47
1385 | 1383,윌,47
1386 | 1384,엌,47
1387 | 1385,앓,47
1388 | 1386,씁,47
1389 | 1387,륭,47
1390 | 1388,W,47
1391 | 1389,쑤,46
1392 | 1390,삥,46
1393 | 1391,돕,46
1394 | 1392,깰,46
1395 | 1393,핍,45
1396 | 1394,텃,45
1397 | 1395,슛,45
1398 | 1396,맸,45
1399 | 1397,롬,45
1400 | 1398,갭,45
1401 | 1399,얽,44
1402 | 1400,쏭,44
1403 | 1401,랠,44
1404 | 1402,겔,44
1405 | 1403,Q,44
1406 | 1404,핥,43
1407 | 1405,킴,43
1408 | 1406,읏,43
1409 | 1407,앚,43
1410 | 1408,숯,43
1411 | 1409,밋,43
1412 | 1410,뽐,42
1413 | 1411,뻣,42
1414 | 1412,눴,42
1415 | 1413,잭,41
1416 | 1414,뽈,41
1417 | 1415,뗐,41
1418 | 1416,꽝,41
1419 | 1417,훑,40
1420 | 1418,캥,40
1421 | 1419,쫀,40
1422 | 1420,뵙,40
1423 | 1421,홋,39
1424 | 1422,펍,39
1425 | 1423,뺨,39
1426 | 1424,됬,39
1427 | 1425,끽,39
1428 | 1426,빕,38
1429 | 1427,밉,38
1430 | 1428,꿋,38
1431 | 1429,헉,37
1432 | 1430,캣,37
1433 | 1431,촘,37
1434 | 1432,셌,37
1435 | 1433,삔,37
1436 | 1434,삑,37
1437 | 1435,뽂,37
1438 | 1436,뮌,37
1439 | 1437,뗄,37
1440 | 1438,텼,36
1441 | 1439,탬,36
1442 | 1440,쨍,36
1443 | 1441,웜,36
1444 | 1442,앰,36
1445 | 1443,맴,36
1446 | 1444,띡,36
1447 | 1445,꿇,36
1448 | 1446,걀,36
1449 | 1447,흩,35
1450 | 1448,쥴,35
1451 | 1449,씸,35
1452 | 1450,낡,35
1453 | 1451,곁,35
1454 | 1452,휙,34
1455 | 1453,쿤,34
1456 | 1454,켄,34
1457 | 1455,츤,34
1458 | 1456,얕,34
1459 | 1457,썪,34
1460 | 1458,둠,34
1461 | 1459,촛,33
1462 | 1460,챌,33
1463 | 1461,죙,33
1464 | 1462,쟈,33
1465 | 1463,잰,33
1466 | 1464,뵀,33
1467 | 1465,w,33
1468 | 1466,푠,32
1469 | 1467,폿,32
1470 | 1468,쨈,32
1471 | 1469,밲,32
1472 | 1470,랖,32
1473 | 1471,떵,32
1474 | 1472,찻,31
1475 | 1473,옇,31
1476 | 1474,뽁,31
1477 | 1475,둣,31
1478 | 1476,닳,31
1479 | 1477,긱,31
1480 | 1478,곶,31
1481 | 1479,휠,30
1482 | 1480,춧,30
1483 | 1481,쐈,30
1484 | 1482,썽,30
1485 | 1483,뼛,29
1486 | 1484,떳,29
1487 | 1485,굘,29
1488 | 1486,훗,28
1489 | 1487,퀵,28
1490 | 1488,봬,28
1491 | 1489,릅,28
1492 | 1490,꺠,28
1493 | 1491,슉,27
1494 | 1492,눔,27
1495 | 1493,끙,27
1496 | 1494,궐,27
1497 | 1495,2,27
1498 | 1496,촥,26
1499 | 1497,젬,26
1500 | 1498,솟,26
1501 | 1499,맷,26
1502 | 1500,룐,26
1503 | 1501,뎃,26
1504 | 1502,깽,26
1505 | 1503,툼,25
1506 | 1504,쎌,25
1507 | 1505,쉼,25
1508 | 1506,쉘,25
1509 | 1507,숑,25
1510 | 1508,뎀,25
1511 | 1509,냔,25
1512 | 1510,쫘,24
1513 | 1511,쎘,24
1514 | 1512,싣,24
1515 | 1513,섣,24
1516 | 1514,샾,24
1517 | 1515,맬,24
1518 | 1516,뗀,24
1519 | 1517,꾀,24
1520 | 1518,헹,23
1521 | 1519,햐,23
1522 | 1520,톰,23
1523 | 1521,췌,23
1524 | 1522,챈,23
1525 | 1523,봔,23
1526 | 1524,밧,23
1527 | 1525,맏,23
1528 | 1526,딥,23
1529 | 1527,늠,23
1530 | 1528,낵,23
1531 | 1529,낱,23
1532 | 1530,꺄,23
1533 | 1531,갬,23
1534 | 1532,훼,22
1535 | 1533,핼,22
1536 | 1534,튠,22
1537 | 1535,웩,22
1538 | 1536,쏜,22
1539 | 1537,뿅,22
1540 | 1538,빰,22
1541 | 1539,딤,22
1542 | 1540,꿉,22
1543 | 1541,걜,22
1544 | 1542,1,22
1545 | 1543,짙,21
1546 | 1544,얍,21
1547 | 1545,샛,21
1548 | 1546,뗘,21
1549 | 1547,듭,21
1550 | 1548,챘,20
1551 | 1549,쯧,20
1552 | 1550,짹,20
1553 | 1551,잦,20
1554 | 1552,옐,20
1555 | 1553,빳,20
1556 | 1554,몹,20
1557 | 1555,몄,20
1558 | 1556,똔,20
1559 | 1557,딧,20
1560 | 1558,놉,20
1561 | 1559,궜,20
1562 | 1560,굼,20
1563 | 1561,헥,19
1564 | 1562,캬,19
1565 | 1563,챕,19
1566 | 1564,쟨,19
1567 | 1565,멓,19
1568 | 1566,똠,19
1569 | 1567,댐,19
1570 | 1568,텁,18
1571 | 1569,켈,18
1572 | 1570,첵,18
1573 | 1571,숄,18
1574 | 1572,띨,18
1575 | 1573,듦,18
1576 | 1574,궤,18
1577 | 1575,곗,18
1578 | 1576,튈,17
1579 | 1577,좆,17
1580 | 1578,윷,17
1581 | 1579,옅,17
1582 | 1580,얏,17
1583 | 1581,믈,17
1584 | 1582,룽,17
1585 | 1583,띃,17
1586 | 1584,딕,17
1587 | 1585,뎁,17
1588 | 1586,닛,17
1589 | 1587,냑,17
1590 | 1588,겅,17
1591 | 1589,휩,16
1592 | 1590,팎,16
1593 | 1591,틋,16
1594 | 1592,콸,16
1595 | 1593,콥,16
1596 | 1594,잴,16
1597 | 1595,웁,16
1598 | 1596,슘,16
1599 | 1597,멱,16
1600 | 1598,랏,16
1601 | 1599,떄,16
1602 | 1600,뒨,16
1603 | 1601,꿰,16
1604 | 1602,깻,16
1605 | 1603,긌,16
1606 | 1604,젼,15
1607 | 1605,윰,15
1608 | 1606,웍,15
1609 | 1607,앳,15
1610 | 1608,샬,15
1611 | 1609,샥,15
1612 | 1610,볕,15
1613 | 1611,멩,15
1614 | 1612,넹,15
1615 | 1613,넙,15
1616 | 1614,끕,15
1617 | 1615,휜,14
1618 | 1616,텄,14
1619 | 1617,쫒,14
1620 | 1618,쩝,14
1621 | 1619,쨋,14
1622 | 1620,윳,14
1623 | 1621,쉭,14
1624 | 1622,쇽,14
1625 | 1623,셧,14
1626 | 1624,뵐,14
1627 | 1625,땔,14
1628 | 1626,덱,14
1629 | 1627,댑,14
1630 | 1628,꺽,14
1631 | 1629,곪,14
1632 | 1630,켔,13
1633 | 1631,츰,13
1634 | 1632,읜,13
1635 | 1633,쑈,13
1636 | 1634,볐,13
1637 | 1635,및,13
1638 | 1636,롷,13
1639 | 1637,딛,13
1640 | 1638,냇,13
1641 | 1639,3,13
1642 | 1640,혓,12
1643 | 1641,팻,12
1644 | 1642,팰,12
1645 | 1643,킁,12
1646 | 1644,촤,12
1647 | 1645,쨰,12
1648 | 1646,잿,12
1649 | 1647,옌,12
1650 | 1648,쐬,12
1651 | 1649,쌋,12
1652 | 1650,슁,12
1653 | 1651,쉑,12
1654 | 1652,빢,12
1655 | 1653,뵌,12
1656 | 1654,뭄,12
1657 | 1655,묽,12
1658 | 1656,뎠,12
1659 | 1657,늪,12
1660 | 1658,뇽,12
1661 | 1659,훤,11
1662 | 1660,횔,11
1663 | 1661,홧,11
1664 | 1662,햅,11
1665 | 1663,푯,11
1666 | 1664,팼,11
1667 | 1665,탯,11
1668 | 1666,탤,11
1669 | 1667,큭,11
1670 | 1668,짢,11
1671 | 1669,읊,11
1672 | 1670,롸,11
1673 | 1671,띤,11
1674 | 1672,놋,11
1675 | 1673,넴,11
1676 | 1674,귿,11
1677 | 1675,q,11
1678 | 1676,휑,10
1679 | 1677,퓸,10
1680 | 1678,튿,10
1681 | 1679,튄,10
1682 | 1680,촐,10
1683 | 1681,쭌,10
1684 | 1682,짊,10
1685 | 1683,숴,10
1686 | 1684,숀,10
1687 | 1685,뿡,10
1688 | 1686,뻬,10
1689 | 1687,렜,10
1690 | 1688,뗬,10
1691 | 1689,늄,10
1692 | 1690,끍,10
1693 | 1691,곌,10
1694 | 1692,갰,10
1695 | 1693,폄,9
1696 | 1694,콰,9
1697 | 1695,캇,9
1698 | 1696,캅,9
1699 | 1697,늉,9
1700 | 1698,뉜,9
1701 | 1699,큔,8
1702 | 1700,콱,8
1703 | 1701,켠,8
1704 | 1702,쳔,8
1705 | 1703,쳇,8
1706 | 1704,읔,8
1707 | 1705,읎,8
1708 | 1706,욤,8
1709 | 1707,뼘,8
1710 | 1708,롹,8
1711 | 1709,렝,8
1712 | 1710,뚠,8
1713 | 1711,땋,8
1714 | 1712,덨,8
1715 | 1713,넵,8
1716 | 1714,넝,8
1717 | 1715,넋,8
1718 | 1716,꿩,8
1719 | 1717,꼿,8
1720 | 1718,깟,8
1721 | 1719,곯,8
1722 | 1720,4,8
1723 | 1721,힝,7
1724 | 1722,헀,7
1725 | 1723,푤,7
1726 | 1724,쿱,7
1727 | 1725,캤,7
1728 | 1726,챗,7
1729 | 1727,쯩,7
1730 | 1728,쮸,7
1731 | 1729,읗,7
1732 | 1730,윅,7
1733 | 1731,얠,7
1734 | 1732,씰,7
1735 | 1733,썅,7
1736 | 1734,쌜,7
1737 | 1735,쌉,7
1738 | 1736,슌,7
1739 | 1737,쉿,7
1740 | 1738,쇳,7
1741 | 1739,셍,7
1742 | 1740,뿍,7
1743 | 1741,뼌,7
1744 | 1742,뺌,7
1745 | 1743,밈,7
1746 | 1744,룟,7
1747 | 1745,룔,7
1748 | 1746,뜀,7
1749 | 1747,끅,7
1750 | 1748,꾜,7
1751 | 1749,겡,7
1752 | 1750,Z,7
1753 | 1751,횰,6
1754 | 1752,헴,6
1755 | 1753,핌,6
1756 | 1754,펩,6
1757 | 1755,펨,6
1758 | 1756,켸,6
1759 | 1757,쩠,6
1760 | 1758,잽,6
1761 | 1759,엡,6
1762 | 1760,앝,6
1763 | 1761,쓕,6
1764 | 1762,썜,6
1765 | 1763,쌘,6
1766 | 1764,삣,6
1767 | 1765,빻,6
1768 | 1766,몀,6
1769 | 1767,뤼,6
1770 | 1768,롄,6
1771 | 1769,떫,6
1772 | 1770,덫,6
1773 | 1771,뉸,6
1774 | 1772,꽥,6
1775 | 1773,궂,6
1776 | 1774,괭,6
1777 | 1775,0,6
1778 | 1776,힉,5
1779 | 1777,휸,5
1780 | 1778,퓰,5
1781 | 1779,퉤,5
1782 | 1780,퀭,5
1783 | 1781,켐,5
1784 | 1782,췻,5
1785 | 1783,챡,5
1786 | 1784,쨀,5
1787 | 1785,젱,5
1788 | 1786,읒,5
1789 | 1787,웡,5
1790 | 1788,옙,5
1791 | 1789,얬,5
1792 | 1790,앎,5
1793 | 1791,씼,5
1794 | 1792,쑨,5
1795 | 1793,쐐,5
1796 | 1794,쌨,5
1797 | 1795,쉈,5
1798 | 1796,숩,5
1799 | 1797,셤,5
1800 | 1798,삠,5
1801 | 1799,뱄,5
1802 | 1800,뭥,5
1803 | 1801,멨,5
1804 | 1802,먀,5
1805 | 1803,랒,5
1806 | 1804,땟,5
1807 | 1805,땁,5
1808 | 1806,뉠,5
1809 | 1807,꽌,5
1810 | 1808,귐,5
1811 | 1809,굥,5
1812 | 1810,굣,5
1813 | 1811,겋,5
1814 | 1812,8,5
1815 | 1813,5,5
1816 | 1814,훙,4
1817 | 1815,혠,4
1818 | 1816,폔,4
1819 | 1817,튬,4
1820 | 1818,퉜,4
1821 | 1819,탉,4
1822 | 1820,큽,4
1823 | 1821,퀀,4
1824 | 1822,쿰,4
1825 | 1823,켤,4
1826 | 1824,켕,4
1827 | 1825,칡,4
1828 | 1826,쯘,4
1829 | 1827,짯,4
1830 | 1828,짔,4
1831 | 1829,쥔,4
1832 | 1830,줜,4
1833 | 1831,죗,4
1834 | 1832,욧,4
1835 | 1833,쒸,4
1836 | 1834,쏼,4
1837 | 1835,쌂,4
1838 | 1836,셴,4
1839 | 1837,샨,4
1840 | 1838,뽜,4
1841 | 1839,뻠,4
1842 | 1840,빝,4
1843 | 1841,봥,4
1844 | 1842,뫼,4
1845 | 1843,맜,4
1846 | 1844,맀,4
1847 | 1845,뤠,4
1848 | 1846,띌,4
1849 | 1847,띈,4
1850 | 1848,뛌,4
1851 | 1849,뚸,4
1852 | 1850,떽,4
1853 | 1851,떰,4
1854 | 1852,떈,4
1855 | 1853,땍,4
1856 | 1854,돠,4
1857 | 1855,뎄,4
1858 | 1856,늣,4
1859 | 1857,뇸,4
1860 | 1858,놥,4
1861 | 1859,녈,4
1862 | 1860,넒,4
1863 | 1861,낏,4
1864 | 1862,갉,4
1865 | 1863,z,4
1866 | 1864,힛,3
1867 | 1865,휀,3
1868 | 1866,헙,3
1869 | 1867,픗,3
1870 | 1868,퐈,3
1871 | 1869,팹,3
1872 | 1870,틔,3
1873 | 1871,퇸,3
1874 | 1872,텟,3
1875 | 1873,탰,3
1876 | 1874,퀼,3
1877 | 1875,칬,3
1878 | 1876,췰,3
1879 | 1877,쳥,3
1880 | 1878,챂,3
1881 | 1879,찧,3
1882 | 1880,쭙,3
1883 | 1881,쬬,3
1884 | 1882,쩄,3
1885 | 1883,쨉,3
1886 | 1884,줴,3
1887 | 1885,죵,3
1888 | 1886,좍,3
1889 | 1887,좃,3
1890 | 1888,읐,3
1891 | 1889,읃,3
1892 | 1890,웟,3
1893 | 1891,왯,3
1894 | 1892,옭,3
1895 | 1893,씅,3
1896 | 1894,쒀,3
1897 | 1895,쑹,3
1898 | 1896,쏵,3
1899 | 1897,쎔,3
1900 | 1898,쎅,3
1901 | 1899,슾,3
1902 | 1900,쉣,3
1903 | 1901,솨,3
1904 | 1902,셸,3
1905 | 1903,셩,3
1906 | 1904,삯,3
1907 | 1905,뿟,3
1908 | 1906,뾱,3
1909 | 1907,뽝,3
1910 | 1908,뻰,3
1911 | 1909,뵤,3
1912 | 1910,볏,3
1913 | 1911,뱌,3
1914 | 1912,뱁,3
1915 | 1913,밨,3
1916 | 1914,밎,3
1917 | 1915,믐,3
1918 | 1916,뭡,3
1919 | 1917,뭠,3
1920 | 1918,뭍,3
1921 | 1919,묭,3
1922 | 1920,뫠,3
1923 | 1921,멏,3
1924 | 1922,멎,3
1925 | 1923,맽,3
1926 | 1924,뙤,3
1927 | 1925,떔,3
1928 | 1926,땃,3
1929 | 1927,덷,3
1930 | 1928,댜,3
1931 | 1929,늰,3
1932 | 1930,놘,3
1933 | 1931,냘,3
1934 | 1932,뀜,3
1935 | 1933,꽹,3
1936 | 1934,꽐,3
1937 | 1935,꼇,3
1938 | 1936,껸,3
1939 | 1937,껨,3
1940 | 1938,꺅,3
1941 | 1939,곘,3
1942 | 1940,겝,3
1943 | 1941,흣,2
1944 | 1942,흝,2
1945 | 1943,흄,2
1946 | 1944,헸,2
1947 | 1945,헜,2
1948 | 1946,햘,2
1949 | 1947,햑,2
1950 | 1948,핟,2
1951 | 1949,픙,2
1952 | 1950,픕,2
1953 | 1951,푀,2
1954 | 1952,퐉,2
1955 | 1953,틑,2
1956 | 1954,틍,2
1957 | 1955,큣,2
1958 | 1956,퀏,2
1959 | 1957,콴,2
1960 | 1958,켁,2
1961 | 1959,칟,2
1962 | 1960,츳,2
1963 | 1961,츨,2
1964 | 1962,츈,2
1965 | 1963,쵝,2
1966 | 1964,촙,2
1967 | 1965,찦,2
1968 | 1966,찟,2
1969 | 1967,쭁,2
1970 | 1968,쫏,2
1971 | 1969,쩬,2
1972 | 1970,쩧,2
1973 | 1971,짆,2
1974 | 1972,쥰,2
1975 | 1973,쥘,2
1976 | 1974,좐,2
1977 | 1975,졍,2
1978 | 1976,쟝,2
1979 | 1977,잧,2
1980 | 1978,잍,2
1981 | 1979,욍,2
1982 | 1980,왬,2
1983 | 1981,옫,2
1984 | 1982,옜,2
1985 | 1983,옉,2
1986 | 1984,엤,2
1987 | 1985,얐,2
1988 | 1986,얉,2
1989 | 1987,쒯,2
1990 | 1988,쑐,2
1991 | 1989,쏸,2
1992 | 1990,썻,2
1993 | 1991,쌕,2
1994 | 1992,싷,2
1995 | 1993,솝,2
1996 | 1994,솎,2
1997 | 1995,샅,2
1998 | 1996,삻,2
1999 | 1997,쁩,2
2000 | 1998,빘,2
2001 | 1999,뷸,2
2002 | 2000,뷴,2
2003 | 2001,봽,2
2004 | 2002,봈,2
2005 | 2003,볌,2
2006 | 2004,벋,2
2007 | 2005,믕,2
2008 | 2006,뮴,2
2009 | 2007,뭬,2
2010 | 2008,뭑,2
2011 | 2009,묜,2
2012 | 2010,맇,2
2013 | 2011,릈,2
2014 | 2012,륑,2
2015 | 2013,랮,2
2016 | 2014,랟,2
2017 | 2015,띔,2
2018 | 2016,떱,2
2019 | 2017,듈,2
2020 | 2018,됴,2
2021 | 2019,됫,2
2022 | 2020,됙,2
2023 | 2021,됑,2
2024 | 2022,댤,2
2025 | 2023,닯,2
2026 | 2024,늗,2
2027 | 2025,뉩,2
2028 | 2026,눟,2
2029 | 2027,눗,2
2030 | 2028,넚,2
2031 | 2029,냡,2
2032 | 2030,낢,2
2033 | 2031,꾿,2
2034 | 2032,꼳,2
2035 | 2033,꼄,2
2036 | 2034,겊,2
2037 | 2035,갼,2
2038 | 2036,6,2
2039 | 2037,<s>,0
2040 | 2038,</s>,0
2041 | 2039,_,0
2042 | 


--------------------------------------------------------------------------------