├── FLUTEfinaltest.json ├── FLUTEfinaltrain.json ├── FLUTEfinalval.json ├── README.md ├── SarcasmNLI └── sarcasm_train.jsonl ├── generateLabel&Explanations.py ├── idiomNLI └── idiom_train.jsonl ├── metaphorNLI └── metaphor_train.jsonl ├── modelpredictions ├── idiom.json ├── metaphor.json ├── sarcasm.json └── simile.json ├── prompts ├── simile_contra_prompt.txt ├── simile_explanation_prompt.txt └── simile_paraphrase_prompt.txt ├── run_translation.py ├── simileNLI └── simile_train.jsonl ├── testgolddata ├── idiom_test.jsonl ├── metaphor_test.jsonl ├── sarcasm_test.jsonl └── simile_test.jsonl └── train_I->OR.sh /README.md: -------------------------------------------------------------------------------- 1 | # model-in-the-loop-fig-lang 2 | 3 | Code and Data for EMNLP 2022 paper FLUTE: Figurative Language Understanding through Textual Explanations
4 | Email : tuhin.chakr@cs.columbia.edu ( For enquiries) 5 | 6 | 7 | 8 | You can explore our data at https://huggingface.co/datasets/ColumbiaNLP/FLUTE with dataset explorer
9 | Our train, val and test data can be accessed from files FLUTEfinaltrain.json,FLUTEfinalval.json,FLUTEfinaltest.json 10 | The test labels and explanations for the entire test set are intentionally hidden for now.We will release it in time.Please email us for gold label and explanations for test set 11 | 12 | To train the model I->OR model which jointly generates label and rationale
13 | 14 | 15 | Download transformers and install it
16 | Replace the run_translation.py file from here in the folder transformers/examples/pytorch/translation/
17 | Run bash train_I->OR.sh 18 | 19 | To get predictions 20 | 21 | Run generateLabel&Explanations.py 22 | 23 | To evaluate using automated metrics use 24 | 25 | 26 | See this colab https://colab.research.google.com/drive/1uV26GNlltIlnzRe_IDvVLSmKQsa7yj-7?usp=sharing 27 | 28 | To access our human eval data look at the folder for humanevaluation 29 | To access our look at modelpredictions folder 30 | 31 | 32 | 33 | @inproceedings{Chakrabarty2022FLUTEFL, 34 | title={FLUTE: Figurative Language Understanding through Textual Explanations}, 35 | author={Tuhin Chakrabarty and A. Saakyan and Debanjan Ghosh and Smaranda Muresan}, 36 | year={2022} 37 | } 38 | -------------------------------------------------------------------------------- /generateLabel&Explanations.py: -------------------------------------------------------------------------------- 1 | import os 2 | import json 3 | os.environ["CUDA_VISIBLE_DEVICES"]="0" 4 | from transformers import AutoTokenizer, AutoModelForSeq2SeqLM 5 | import numpy as np 6 | import random 7 | import torch 8 | torch.manual_seed(0) 9 | random.seed(0) 10 | np.random.seed(0) 11 | 12 | 13 | data = [] 14 | for line in open("test.jsonl"): 15 | line = json.loads(line) 16 | data.append(line) 17 | 18 | model_dir = "./output/" 19 | tokenizer = AutoTokenizer.from_pretrained(model_dir) 20 | model = AutoModelForSeq2SeqLM.from_pretrained(model_dir) 21 | model.cuda() 22 | 23 | 24 | for i in range(len(data)): 25 | sentence_A = data[i]["premise"] 26 | sentence_B = data[i]["hypothesis"] 27 | instruction = "Does the sentence "+'"'+sentence_A+'" entail or contradict the sentence "'+sentence_B+'"? Please answer between '+'"Entails" or "Contradicts" and explain your decision in a sentence.' 28 | inputs = tokenizer.encode(instruction, return_tensors="pt") 29 | outputs = model.generate(input_ids=inputs.cuda(), no_repeat_ngram_size=2, min_length=10, do_sample=True, max_length=128, top_k=5,temperature=0.7,eos_token_id=tokenizer.eos_token_id) 30 | answer = tokenizer.decode(outputs[0],skip_special_tokens=True) 31 | if "Entails." in answer: 32 | predictedlabel = "Entailment" 33 | predictedExpl = answer.split("Entails.")[1].lstrip().capitalize() 34 | elif "Contradicts." in answer: 35 | predictedlabel = "Contradiction" 36 | predictedExpl = answer.split("Contradicts.")[1].lstrip().capitalize() 37 | data[i]["predicted_label"] = predictedlabel 38 | data[i]["model_explanation"] = predictedExpl 39 | 40 | with open("./output/predictions.json","w") as f: 41 | f.write(json.dumps(data,indent=4)) 42 | 43 | 44 | -------------------------------------------------------------------------------- /prompts/simile_contra_prompt.txt: -------------------------------------------------------------------------------- 1 | You will be presented with a Sentence and a Property. Your goal is to invert the meaning of the Sentence with respect to the Property via a minimal edit. 2 | 3 | 1. Sentence: The place looked impenetrable and inescapable 4 | Property: impenetrable and inescapable 5 | Inversion: This place looked easy to walk into and exit from 6 | 7 | 2. Sentence: He was looking confident. 8 | Property: confident 9 | Inversion: He was looking insecure. 10 | 11 | 3. Sentence: The air inside the pipe was cool and musty. 12 | Property: cool and musty 13 | Inversion: The air inside the pipe was hot and fresh. 14 | 15 | 4. Sentence: I was tired and strengthless when kris got back home at spot-on ten to eight. 16 | Property: tired and strengthless 17 | Inversion: I was energized and invigorated when kris got back home at spot-on ten to eight. 18 | 19 | 5. Sentence: They were crystal blue. 20 | Property: crystal blue 21 | Inversion: They were murky and dark. 22 | 23 | 6. Sentence: The voice sounded strange and robotic on the other end of the phone line. 24 | Property: strange, robotic 25 | Inversion: The voice sounded natural and human on the other end of the phone line.. 26 | 27 | 7. Sentence: {sentence} 28 | Property: {prop} 29 | Inversion: -------------------------------------------------------------------------------- /prompts/simile_explanation_prompt.txt: -------------------------------------------------------------------------------- 1 | You will be provided with a Simile and a contradictory sentence after the word "Contradiction". Your task is to explain why the contradictory sentence contradicts the Simile. 2 | 3 | 1. Simile: like a psychic whirlpool , it began to draw him in. 4 | Contradiction: Mildly, it began to draw him in 5 | Explanation: A whirlpool is a strong current, so a psychic whirlpool drawing in indicates that it was drawing him in intensely, rather than mildly. 6 | 7 | 2. Simile: Her short, blue hair stuck out all over the place, her eyes were like a goth clown . 8 | Contradiction: She had eyes that looked healthy and clear, and her hair was blue, cut short, and messy. 9 | Explanation: Goth clowns typically have lots of make up, including using red color, that might make it look like her eyes are bloodshot, and not healthy. 10 | 11 | 3. Simile: The place looked like a fortress 12 | Contradiction: The place looked easy to walk into and exit from 13 | Explanation: A fortress is a military stronghold, hence it would be very hard to walk into and exit from it. 14 | 15 | 4. Simile: He was looking like a true rajput warrior . 16 | Contradiction: He was looking underconfident. 17 | Explanation: A true warrior has to demonstrate qualities like confidence and assertiveness, hence an underconfident person would not be looking like a true warrior. 18 | 19 | 5. Simile: As he dangled there, fighting for a toehold, his horse went like a punting boat in the wake of a royal barge. 20 | Contradiction: As the racehorse went swiftly forward, he was on the verge of falling from it. 21 | Explanation: A royal barge is a large, threatening vessel, whereas a punting boat is a small vessel that will go astern to avoid confrontation with the barge, not forward. 22 | 23 | 6. Simile: She looked like a broken china doll . 24 | Contradiction: She looked so strong. 25 | Explanation: A china doll is made out of porcelain, which is a fragile material, hence, looking like her could not be perceived as strong. 26 | 27 | 7. Simile: It was like a woodland deadfall in winter before the snow. 28 | Contradiction: It was warm and humid 29 | Explanation: A deadfall is a mass of fallen trees and branches, so saying something is like a deadfall implies it is dry and not humid, and it is cold and not warm in winter. 30 | 31 | 8. Simile: {simile} 32 | Contradiction: {contra} 33 | Explanation: -------------------------------------------------------------------------------- /prompts/simile_paraphrase_prompt.txt: -------------------------------------------------------------------------------- 1 | You will be presented with examples of some literal input sentences and their creative paraphrases. You will also be presented words that need to be preserved. Your task is to generate a creative paraphrase for a given literal sentence consistent in meaning. DO NOT CHANGE words after "Preserve:" keyword. 2 | 3 | 1. Sentence: overwhelmingly , it began to draw him in. 4 | Preserve: overwhelmingly 5 | Creative Paraphrase: He was overwhelmingly obsessed with it. 6 | 7 | 2. Sentence: Her short, blue hair stuck out all over the place, her eyes were bloodshot and dripped black . 8 | Preserve: bloodshot and dripped back 9 | Creative Paraphrase: She had eyes that looked bloodshot and dripped back, and her hair was dark blue, cut short, and messy. 10 | 11 | 3. Sentence: The place looked impenetrable and inescapable 12 | Preserve: impenetrable and inescapable 13 | Creative Paraphrase: The area seemed impenetrable and inescapable 14 | 15 | 4. Sentence: As he dangled there, fighting for a toehold, his horse went swiftly astern 16 | Preserve: swiftly astern 17 | Creative Paraphrase: As the racehorse went swiftly astern, he was on the verge of falling from it 18 | 19 | 5. Sentence: Scattered sunlight poked through the leaves diaphanously . 20 | Preserve: diaphanously 21 | Creative Paraphrase: Rays of sun diaphanously pierced through the plants 22 | 23 | 6. Sentence: The khan motioned widely towards alejo, pleased 24 | Preserve: pleased 25 | Creative Paraphrase: The tsar was pleased as he moved closer to him. 26 | 27 | 7. Sentence: It was becoming nothing but a sliver on the horizon growing smaller and smaller . 28 | Preserve: smaller and smaller 29 | Creative Paraphrase: It was rapidly becoming smaller and smaller on the horizon, turning into nothing but a silver. 30 | 31 | 8. Sentence: Her grey blue eyes gaze at the three men became fierce . 32 | Preserve: fierce 33 | Creative Paraphrase: Her blue-grey eyes became fierce as she stared at the three humans. 34 | 35 | 9. Sentence: He cruised slowly down main street, which was almost entirely empty . 36 | Preserve: entirely empty 37 | Creative Paraphrase: The only thing walking down main street was him, as it was entirely empty. 38 | 39 | 10. Sentence: Really, whatever part of him was decently cute or hot has suddenly become putrid . 40 | Preserve: putrid 41 | Creative Paraphrase: Whatever part of him that was remotely attractive has now become putrid. 42 | 43 | 14. Sentence: The fall colors were rich and striking, but even that didn't lift her spirits. 44 | Preserve: rich and striking 45 | Creative Paraphrase: Even though the fall colors were rich and striking, her spirits were still low. 46 | 47 | 15. Sentence: The taste of his mouth on my lips was sweet 48 | Preserve: sweet 49 | Creative Paraphrase: His mouth tasted sweet when he kissed me. 50 | 51 | 16. Sentence: {sentence} 52 | Preserve: {preserve} 53 | Creative Paraphrase: -------------------------------------------------------------------------------- /run_translation.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | # coding=utf-8 3 | # Copyright The HuggingFace Team and The HuggingFace Inc. team. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | """ 17 | Fine-tuning the library models for sequence to sequence. 18 | """ 19 | # You can also adapt this script on your own sequence to sequence task. Pointers for this are left as comments. 20 | 21 | import logging 22 | import os 23 | import sys 24 | from dataclasses import dataclass, field 25 | from typing import Optional 26 | 27 | import datasets 28 | import numpy as np 29 | from datasets import load_dataset, load_metric 30 | os.environ["CUDA_VISIBLE_DEVICES"]="0,1,2,3" 31 | 32 | import transformers 33 | from transformers import ( 34 | AutoConfig, 35 | AutoModelForSeq2SeqLM, 36 | AutoTokenizer, 37 | DataCollatorForSeq2Seq, 38 | HfArgumentParser, 39 | M2M100Tokenizer, 40 | MBart50Tokenizer, 41 | MBart50TokenizerFast, 42 | MBartTokenizer, 43 | MBartTokenizerFast, 44 | Seq2SeqTrainer, 45 | Seq2SeqTrainingArguments, 46 | default_data_collator, 47 | set_seed, 48 | ) 49 | from transformers.trainer_utils import get_last_checkpoint 50 | from transformers.utils import check_min_version 51 | from transformers.utils.versions import require_version 52 | import torch 53 | import pdb 54 | torch.backends.cuda.matmul.allow_tf32 = True 55 | 56 | 57 | # Will error if the minimal version of Transformers is not installed. Remove at your own risks. 58 | check_min_version("4.13.0.dev0") 59 | 60 | require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/translation/requirements.txt") 61 | 62 | logger = logging.getLogger(__name__) 63 | 64 | # A list of all multilingual tokenizer which require src_lang and tgt_lang attributes. 65 | MULTILINGUAL_TOKENIZERS = [MBartTokenizer, MBartTokenizerFast, MBart50Tokenizer, MBart50TokenizerFast, M2M100Tokenizer] 66 | 67 | 68 | @dataclass 69 | class ModelArguments: 70 | """ 71 | Arguments pertaining to which model/config/tokenizer we are going to fine-tune from. 72 | """ 73 | 74 | model_name_or_path: str = field( 75 | metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"} 76 | ) 77 | config_name: Optional[str] = field( 78 | default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"} 79 | ) 80 | tokenizer_name: Optional[str] = field( 81 | default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"} 82 | ) 83 | cache_dir: Optional[str] = field( 84 | default=None, 85 | metadata={"help": "Where to store the pretrained models downloaded from huggingface.co"}, 86 | ) 87 | use_fast_tokenizer: bool = field( 88 | default=True, 89 | metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."}, 90 | ) 91 | model_revision: str = field( 92 | default="main", 93 | metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."}, 94 | ) 95 | use_auth_token: bool = field( 96 | default=False, 97 | metadata={ 98 | "help": "Will use the token generated when running `transformers-cli login` (necessary to use this script " 99 | "with private models)." 100 | }, 101 | ) 102 | 103 | 104 | @dataclass 105 | class DataTrainingArguments: 106 | """ 107 | Arguments pertaining to what data we are going to input our model for training and eval. 108 | """ 109 | 110 | source_lang: str = field(default=None, metadata={"help": "Source language id for translation."}) 111 | target_lang: str = field(default=None, metadata={"help": "Target language id for translation."}) 112 | 113 | dataset_name: Optional[str] = field( 114 | default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."} 115 | ) 116 | dataset_config_name: Optional[str] = field( 117 | default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."} 118 | ) 119 | train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a jsonlines)."}) 120 | validation_file: Optional[str] = field( 121 | default=None, 122 | metadata={ 123 | "help": "An optional input evaluation data file to evaluate the metrics (sacreblue) on " 124 | "a jsonlines file." 125 | }, 126 | ) 127 | test_file: Optional[str] = field( 128 | default=None, 129 | metadata={ 130 | "help": "An optional input test data file to evaluate the metrics (sacreblue) on " "a jsonlines file." 131 | }, 132 | ) 133 | overwrite_cache: bool = field( 134 | default=False, metadata={"help": "Overwrite the cached training and evaluation sets"} 135 | ) 136 | preprocessing_num_workers: Optional[int] = field( 137 | default=None, 138 | metadata={"help": "The number of processes to use for the preprocessing."}, 139 | ) 140 | max_source_length: Optional[int] = field( 141 | default=1024, 142 | metadata={ 143 | "help": "The maximum total input sequence length after tokenization. Sequences longer " 144 | "than this will be truncated, sequences shorter will be padded." 145 | }, 146 | ) 147 | max_target_length: Optional[int] = field( 148 | default=128, 149 | metadata={ 150 | "help": "The maximum total sequence length for target text after tokenization. Sequences longer " 151 | "than this will be truncated, sequences shorter will be padded." 152 | }, 153 | ) 154 | val_max_target_length: Optional[int] = field( 155 | default=None, 156 | metadata={ 157 | "help": "The maximum total sequence length for validation target text after tokenization. Sequences longer " 158 | "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`." 159 | "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used " 160 | "during ``evaluate`` and ``predict``." 161 | }, 162 | ) 163 | pad_to_max_length: bool = field( 164 | default=False, 165 | metadata={ 166 | "help": "Whether to pad all samples to model maximum sentence length. " 167 | "If False, will pad the samples dynamically when batching to the maximum length in the batch. More " 168 | "efficient on GPU but very bad for TPU." 169 | }, 170 | ) 171 | max_train_samples: Optional[int] = field( 172 | default=None, 173 | metadata={ 174 | "help": "For debugging purposes or quicker training, truncate the number of training examples to this " 175 | "value if set." 176 | }, 177 | ) 178 | max_eval_samples: Optional[int] = field( 179 | default=None, 180 | metadata={ 181 | "help": "For debugging purposes or quicker training, truncate the number of evaluation examples to this " 182 | "value if set." 183 | }, 184 | ) 185 | max_predict_samples: Optional[int] = field( 186 | default=None, 187 | metadata={ 188 | "help": "For debugging purposes or quicker training, truncate the number of prediction examples to this " 189 | "value if set." 190 | }, 191 | ) 192 | num_beams: Optional[int] = field( 193 | default=None, 194 | metadata={ 195 | "help": "Number of beams to use for evaluation. This argument will be passed to ``model.generate``, " 196 | "which is used during ``evaluate`` and ``predict``." 197 | }, 198 | ) 199 | ignore_pad_token_for_loss: bool = field( 200 | default=True, 201 | metadata={ 202 | "help": "Whether to ignore the tokens corresponding to padded labels in the loss computation or not." 203 | }, 204 | ) 205 | source_prefix: Optional[str] = field( 206 | default=None, metadata={"help": "A prefix to add before every source text (useful for T5 models)."} 207 | ) 208 | forced_bos_token: Optional[str] = field( 209 | default=None, 210 | metadata={ 211 | "help": "The token to force as the first generated token after the :obj:`decoder_start_token_id`." 212 | "Useful for multilingual models like :doc:`mBART <../model_doc/mbart>` where the first generated token " 213 | "needs to be the target language token.(Usually it is the target language token)" 214 | }, 215 | ) 216 | 217 | def __post_init__(self): 218 | if self.dataset_name is None and self.train_file is None and self.validation_file is None: 219 | raise ValueError("Need either a dataset name or a training/validation file.") 220 | elif self.source_lang is None or self.target_lang is None: 221 | raise ValueError("Need to specify the source language and the target language.") 222 | 223 | if self.train_file is not None: 224 | extension = self.train_file.split(".")[-1] 225 | assert extension == "json", "`train_file` should be a json file." 226 | if self.validation_file is not None: 227 | extension = self.validation_file.split(".")[-1] 228 | assert extension == "json", "`validation_file` should be a json file." 229 | if self.val_max_target_length is None: 230 | self.val_max_target_length = self.max_target_length 231 | 232 | 233 | def main(): 234 | # See all possible arguments in src/transformers/training_args.py 235 | # or by passing the --help flag to this script. 236 | # We now keep distinct sets of args, for a cleaner separation of concerns. 237 | 238 | parser = HfArgumentParser((ModelArguments, DataTrainingArguments, Seq2SeqTrainingArguments)) 239 | if len(sys.argv) == 2 and sys.argv[1].endswith(".json"): 240 | # If we pass only one argument to the script and it's the path to a json file, 241 | # let's parse it to get our arguments. 242 | model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1])) 243 | else: 244 | model_args, data_args, training_args = parser.parse_args_into_dataclasses() 245 | 246 | logging.basicConfig( 247 | format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", 248 | datefmt="%m/%d/%Y %H:%M:%S", 249 | handlers=[logging.StreamHandler(sys.stdout)], 250 | ) 251 | 252 | log_level = training_args.get_process_log_level() 253 | logger.setLevel(log_level) 254 | datasets.utils.logging.set_verbosity(log_level) 255 | transformers.utils.logging.set_verbosity(log_level) 256 | transformers.utils.logging.enable_default_handler() 257 | transformers.utils.logging.enable_explicit_format() 258 | training_args.lr_scheduler_type = "constant" 259 | 260 | # Log on each process the small summary: 261 | logger.warning( 262 | f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}" 263 | + f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}" 264 | ) 265 | logger.info(f"Training/evaluation parameters {training_args}") 266 | 267 | if data_args.source_prefix is None and model_args.model_name_or_path in [ 268 | "t5-small", 269 | "t5-base", 270 | "t5-large", 271 | "t5-3b", 272 | "t5-11b", 273 | ]: 274 | logger.warning( 275 | "You're running a t5 model but didn't provide a source prefix, which is expected, e.g. with " 276 | "`--source_prefix 'translate English to German: ' `" 277 | ) 278 | 279 | # Detecting last checkpoint. 280 | last_checkpoint = None 281 | if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir: 282 | last_checkpoint = get_last_checkpoint(training_args.output_dir) 283 | if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0: 284 | raise ValueError( 285 | f"Output directory ({training_args.output_dir}) already exists and is not empty. " 286 | "Use --overwrite_output_dir to overcome." 287 | ) 288 | elif last_checkpoint is not None and training_args.resume_from_checkpoint is None: 289 | logger.info( 290 | f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change " 291 | "the `--output_dir` or add `--overwrite_output_dir` to train from scratch." 292 | ) 293 | 294 | # Set seed before initializing model. 295 | set_seed(training_args.seed) 296 | 297 | # Get the datasets: you can either provide your own JSON training and evaluation files (see below) 298 | # or just provide the name of one of the public datasets available on the hub at https://huggingface.co/datasets/ 299 | # (the dataset will be downloaded automatically from the datasets Hub). 300 | # 301 | # For translation, only JSON files are supported, with one field named "translation" containing two keys for the 302 | # source and target languages (unless you adapt what follows). 303 | # 304 | # In distributed training, the load_dataset function guarantee that only one local process can concurrently 305 | # download the dataset. 306 | if data_args.dataset_name is not None: 307 | # Downloading and loading a dataset from the hub. 308 | raw_datasets = load_dataset( 309 | data_args.dataset_name, data_args.dataset_config_name, cache_dir=model_args.cache_dir 310 | ) 311 | else: 312 | data_files = {} 313 | if data_args.train_file is not None: 314 | data_files["train"] = data_args.train_file 315 | extension = data_args.train_file.split(".")[-1] 316 | if data_args.validation_file is not None: 317 | data_files["validation"] = data_args.validation_file 318 | extension = data_args.validation_file.split(".")[-1] 319 | if data_args.test_file is not None: 320 | data_files["test"] = data_args.test_file 321 | extension = data_args.test_file.split(".")[-1] 322 | raw_datasets = load_dataset(extension, data_files=data_files, cache_dir=model_args.cache_dir) 323 | # See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at 324 | # https://huggingface.co/docs/datasets/loading_datasets.html. 325 | 326 | # Load pretrained model and tokenizer 327 | # 328 | # Distributed training: 329 | # The .from_pretrained methods guarantee that only one local process can concurrently 330 | # download model & vocab. 331 | cachemachine = model_args.cache_dir 332 | config = AutoConfig.from_pretrained( 333 | model_args.config_name if model_args.config_name else model_args.model_name_or_path, 334 | cache_dir=cachemachine, 335 | revision=model_args.model_revision, 336 | use_auth_token=True if model_args.use_auth_token else None, 337 | ) 338 | tokenizer = AutoTokenizer.from_pretrained( 339 | model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path, 340 | cache_dir=cachemachine, 341 | use_fast=model_args.use_fast_tokenizer, 342 | revision=model_args.model_revision, 343 | use_auth_token=True if model_args.use_auth_token else None, 344 | ) 345 | model = AutoModelForSeq2SeqLM.from_pretrained( 346 | model_args.model_name_or_path, 347 | from_tf=bool(".ckpt" in model_args.model_name_or_path), 348 | config=config, 349 | cache_dir=cachemachine, 350 | revision=model_args.model_revision, 351 | use_auth_token=True if model_args.use_auth_token else None, 352 | ) 353 | 354 | device_ids = [0,1,2,3] 355 | device_map = {device_ids[0]: list(range(0, 6)),device_ids[1]: list(range(6,12)),device_ids[2]: list(range(12,18)),device_ids[3]: list(range(18,24))} 356 | model.parallelize(device_map) 357 | 358 | model.resize_token_embeddings(len(tokenizer)) 359 | 360 | # Set decoder_start_token_id 361 | if model.config.decoder_start_token_id is None and isinstance(tokenizer, (MBartTokenizer, MBartTokenizerFast)): 362 | if isinstance(tokenizer, MBartTokenizer): 363 | model.config.decoder_start_token_id = tokenizer.lang_code_to_id[data_args.target_lang] 364 | else: 365 | model.config.decoder_start_token_id = tokenizer.convert_tokens_to_ids(data_args.target_lang) 366 | 367 | if model.config.decoder_start_token_id is None: 368 | raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined") 369 | 370 | prefix = data_args.source_prefix if data_args.source_prefix is not None else "" 371 | 372 | # Preprocessing the datasets. 373 | # We need to tokenize inputs and targets. 374 | if training_args.do_train: 375 | column_names = raw_datasets["train"].column_names 376 | elif training_args.do_eval: 377 | column_names = raw_datasets["validation"].column_names 378 | elif training_args.do_predict: 379 | column_names = raw_datasets["test"].column_names 380 | else: 381 | logger.info("There is nothing to do. Please pass `do_train`, `do_eval` and/or `do_predict`.") 382 | return 383 | 384 | # For translation we set the codes of our source and target languages (only useful for mBART, the others will 385 | # ignore those attributes). 386 | if isinstance(tokenizer, tuple(MULTILINGUAL_TOKENIZERS)): 387 | assert data_args.target_lang is not None and data_args.source_lang is not None, ( 388 | f"{tokenizer.__class__.__name__} is a multilingual tokenizer which requires --source_lang and " 389 | "--target_lang arguments." 390 | ) 391 | 392 | tokenizer.src_lang = data_args.source_lang 393 | tokenizer.tgt_lang = data_args.target_lang 394 | 395 | # For multilingual translation models like mBART-50 and M2M100 we need to force the target language token 396 | # as the first generated token. We ask the user to explicitly provide this as --forced_bos_token argument. 397 | forced_bos_token_id = ( 398 | tokenizer.lang_code_to_id[data_args.forced_bos_token] if data_args.forced_bos_token is not None else None 399 | ) 400 | model.config.forced_bos_token_id = forced_bos_token_id 401 | 402 | # Get the language codes for input/target. 403 | source_lang = data_args.source_lang.split("_")[0] 404 | target_lang = data_args.target_lang.split("_")[0] 405 | 406 | # Temporarily set max_target_length for training. 407 | max_target_length = data_args.max_target_length 408 | padding = "max_length" if data_args.pad_to_max_length else False 409 | 410 | if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"): 411 | logger.warning( 412 | "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for" 413 | f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory" 414 | ) 415 | 416 | def preprocess_function(examples): 417 | inputs = [ex["en1"] for ex in examples["translation"]] 418 | targets = [ex["en2"] for ex in examples["translation"]] 419 | inputs = [prefix + inp for inp in inputs] 420 | model_inputs = tokenizer(inputs, max_length=data_args.max_source_length, padding=padding, truncation=True) 421 | 422 | # Setup the tokenizer for targets 423 | with tokenizer.as_target_tokenizer(): 424 | labels = tokenizer(targets, max_length=max_target_length, padding=padding, truncation=True) 425 | 426 | # If we are padding here, replace all tokenizer.pad_token_id in the labels by -100 when we want to ignore 427 | # padding in the loss. 428 | if padding == "max_length" and data_args.ignore_pad_token_for_loss: 429 | labels["input_ids"] = [ 430 | [(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"] 431 | ] 432 | 433 | model_inputs["labels"] = labels["input_ids"] 434 | return model_inputs 435 | 436 | if training_args.do_train: 437 | if "train" not in raw_datasets: 438 | raise ValueError("--do_train requires a train dataset") 439 | train_dataset = raw_datasets["train"] 440 | if data_args.max_train_samples is not None: 441 | train_dataset = train_dataset.select(range(data_args.max_train_samples)) 442 | with training_args.main_process_first(desc="train dataset map pre-processing"): 443 | train_dataset = train_dataset.map( 444 | preprocess_function, 445 | batched=True, 446 | num_proc=data_args.preprocessing_num_workers, 447 | remove_columns=column_names, 448 | load_from_cache_file=not data_args.overwrite_cache, 449 | desc="Running tokenizer on train dataset", 450 | ) 451 | 452 | if training_args.do_eval: 453 | max_target_length = data_args.val_max_target_length 454 | if "validation" not in raw_datasets: 455 | raise ValueError("--do_eval requires a validation dataset") 456 | eval_dataset = raw_datasets["validation"] 457 | if data_args.max_eval_samples is not None: 458 | eval_dataset = eval_dataset.select(range(data_args.max_eval_samples)) 459 | with training_args.main_process_first(desc="validation dataset map pre-processing"): 460 | eval_dataset = eval_dataset.map( 461 | preprocess_function, 462 | batched=True, 463 | num_proc=data_args.preprocessing_num_workers, 464 | remove_columns=column_names, 465 | load_from_cache_file=not data_args.overwrite_cache, 466 | desc="Running tokenizer on validation dataset", 467 | ) 468 | 469 | if training_args.do_predict: 470 | max_target_length = data_args.val_max_target_length 471 | if "test" not in raw_datasets: 472 | raise ValueError("--do_predict requires a test dataset") 473 | predict_dataset = raw_datasets["test"] 474 | if data_args.max_predict_samples is not None: 475 | predict_dataset = predict_dataset.select(range(data_args.max_predict_samples)) 476 | with training_args.main_process_first(desc="prediction dataset map pre-processing"): 477 | predict_dataset = predict_dataset.map( 478 | preprocess_function, 479 | batched=True, 480 | num_proc=data_args.preprocessing_num_workers, 481 | remove_columns=column_names, 482 | load_from_cache_file=not data_args.overwrite_cache, 483 | desc="Running tokenizer on prediction dataset", 484 | ) 485 | 486 | # Data collator 487 | label_pad_token_id = -100 if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id 488 | if data_args.pad_to_max_length: 489 | data_collator = default_data_collator 490 | else: 491 | data_collator = DataCollatorForSeq2Seq( 492 | tokenizer, 493 | model=model, 494 | label_pad_token_id=label_pad_token_id, 495 | pad_to_multiple_of=8 if training_args.fp16 else None, 496 | ) 497 | 498 | # Metric 499 | metric = load_metric("sacrebleu") 500 | 501 | def postprocess_text(preds, labels): 502 | preds = [pred.strip() for pred in preds] 503 | labels = [[label.strip()] for label in labels] 504 | 505 | return preds, labels 506 | 507 | def compute_metrics(eval_preds): 508 | preds, labels = eval_preds 509 | if isinstance(preds, tuple): 510 | preds = preds[0] 511 | decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True) 512 | if data_args.ignore_pad_token_for_loss: 513 | # Replace -100 in the labels as we can't decode them. 514 | labels = np.where(labels != -100, labels, tokenizer.pad_token_id) 515 | decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True) 516 | 517 | # Some simple post-processing 518 | decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels) 519 | 520 | result = metric.compute(predictions=decoded_preds, references=decoded_labels) 521 | result = {"bleu": result["score"]} 522 | 523 | prediction_lens = [np.count_nonzero(pred != tokenizer.pad_token_id) for pred in preds] 524 | result["gen_len"] = np.mean(prediction_lens) 525 | result = {k: round(v, 4) for k, v in result.items()} 526 | return result 527 | 528 | # Initialize our Trainer 529 | trainer = Seq2SeqTrainer( 530 | model=model, 531 | args=training_args, 532 | train_dataset=train_dataset if training_args.do_train else None, 533 | eval_dataset=eval_dataset if training_args.do_eval else None, 534 | tokenizer=tokenizer, 535 | data_collator=data_collator, 536 | compute_metrics=compute_metrics if training_args.predict_with_generate else None, 537 | ) 538 | 539 | # Training 540 | if training_args.do_train: 541 | #pdb.set_trace() 542 | print(train_dataset[0:10]) 543 | checkpoint = None 544 | if training_args.resume_from_checkpoint is not None: 545 | checkpoint = training_args.resume_from_checkpoint 546 | elif last_checkpoint is not None: 547 | checkpoint = last_checkpoint 548 | train_result = trainer.train(resume_from_checkpoint=checkpoint) 549 | trainer.save_model() # Saves the tokenizer too for easy upload 550 | 551 | metrics = train_result.metrics 552 | max_train_samples = ( 553 | data_args.max_train_samples if data_args.max_train_samples is not None else len(train_dataset) 554 | ) 555 | metrics["train_samples"] = min(max_train_samples, len(train_dataset)) 556 | 557 | trainer.log_metrics("train", metrics) 558 | trainer.save_metrics("train", metrics) 559 | trainer.save_state() 560 | 561 | # Evaluation 562 | results = {} 563 | max_length = ( 564 | training_args.generation_max_length 565 | if training_args.generation_max_length is not None 566 | else data_args.val_max_target_length 567 | ) 568 | num_beams = data_args.num_beams if data_args.num_beams is not None else training_args.generation_num_beams 569 | if training_args.do_eval: 570 | logger.info("*** Evaluate ***") 571 | 572 | metrics = trainer.evaluate(max_length=max_length, num_beams=num_beams, metric_key_prefix="eval") 573 | max_eval_samples = data_args.max_eval_samples if data_args.max_eval_samples is not None else len(eval_dataset) 574 | metrics["eval_samples"] = min(max_eval_samples, len(eval_dataset)) 575 | 576 | trainer.log_metrics("eval", metrics) 577 | trainer.save_metrics("eval", metrics) 578 | 579 | if training_args.do_predict: 580 | logger.info("*** Predict ***") 581 | 582 | predict_results = trainer.predict( 583 | predict_dataset, metric_key_prefix="predict", max_length=max_length, num_beams=num_beams 584 | ) 585 | metrics = predict_results.metrics 586 | max_predict_samples = ( 587 | data_args.max_predict_samples if data_args.max_predict_samples is not None else len(predict_dataset) 588 | ) 589 | metrics["predict_samples"] = min(max_predict_samples, len(predict_dataset)) 590 | 591 | trainer.log_metrics("predict", metrics) 592 | trainer.save_metrics("predict", metrics) 593 | 594 | if trainer.is_world_process_zero(): 595 | if training_args.predict_with_generate: 596 | predictions = tokenizer.batch_decode( 597 | predict_results.predictions, skip_special_tokens=True, clean_up_tokenization_spaces=True 598 | ) 599 | predictions = [pred.strip() for pred in predictions] 600 | output_prediction_file = os.path.join(training_args.output_dir, "generated_predictions.txt") 601 | with open(output_prediction_file, "w", encoding="utf-8") as writer: 602 | writer.write("\n".join(predictions)) 603 | 604 | kwargs = {"finetuned_from": model_args.model_name_or_path, "tasks": "translation"} 605 | if data_args.dataset_name is not None: 606 | kwargs["dataset_tags"] = data_args.dataset_name 607 | if data_args.dataset_config_name is not None: 608 | kwargs["dataset_args"] = data_args.dataset_config_name 609 | kwargs["dataset"] = f"{data_args.dataset_name} {data_args.dataset_config_name}" 610 | else: 611 | kwargs["dataset"] = data_args.dataset_name 612 | 613 | languages = [l for l in [data_args.source_lang, data_args.target_lang] if l is not None] 614 | if len(languages) > 0: 615 | kwargs["language"] = languages 616 | 617 | if training_args.push_to_hub: 618 | trainer.push_to_hub(**kwargs) 619 | else: 620 | trainer.create_model_card(**kwargs) 621 | 622 | return results 623 | 624 | 625 | def _mp_fn(index): 626 | # For xla_spawn (TPUs) 627 | main() 628 | 629 | 630 | if __name__ == "__main__": 631 | main() 632 | -------------------------------------------------------------------------------- /testgolddata/metaphor_test.jsonl: -------------------------------------------------------------------------------- 1 | {"id": "1", "hypothesis": "It is sad to observe the fruits of ignorance.", "label": "Entailment", "premise": "It is sad to observe the consequences of ignorance.", "explanation": "The fruits of ignorance means the bad things that happen because someone is ignorant."} 2 | {"id": "1", "hypothesis": "It is sad to observe the fruits of ignorance.", "label": "Contradiction", "premise": "It is amusing to observe the fruits of ignorance.", "explanation": "It is sad to observe the fruits of ignorance if the ignorance is leading to negative consequences, such as people making poor decisions, thus, it can not be amusing to observe the fruits of ignorance at the same time."} 3 | {"id": "2", "hypothesis": "This young man knows how to climb the social ladder.", "label": "Entailment", "premise": "That guy knows how to charm and manage people and work his way up the social ladder.", "explanation": "Climbing the social ladder means advancing in social status."} 4 | {"id": "2", "hypothesis": "This young man knows how to climb the social ladder.", "label": "Contradiction", "premise": "This young man is stuck in the middle of this social ladder.", "explanation": "Climbing the social ladder is a metaphor for improving one's social status while stuck in the middle is a metaphor for being stuck in a difficult or challenging situation with no clear way out."} 5 | {"id": "3", "hypothesis": "The guy was a floating whale in the small swimming pool.", "label": "Entailment", "premise": "A man who is fat and large floats like a log in the small swimming pool.", "explanation": "Whales are very large animals and they float in the water, here, this man is large and fat like a whale and he is floating in the small swimming pool."} 6 | {"id": "3", "hypothesis": "The guy was a floating whale in the small swimming pool.", "label": "Contradiction", "premise": "The guy was a drowning otter in the large swimming pool.", "explanation": "One sentence says that the guy was very large and could not fit in the small pool, whereas the other sentence states that he was a drowning otter in the large swimming pool, i.e. he was small and could not swim in the large pool."} 7 | {"id": "4", "hypothesis": "Was it even now shadowing them, moving soundlessly from cover to cover, like a tiger in the steel jungle ?", "label": "Entailment", "premise": "The fact that I was hiding in the steel jungle, moving from one cover to the other, was like being a tiger.", "explanation": "Being a tiger in the steel jungle means being a predator who is stalking its prey."} 8 | {"id": "4", "hypothesis": "Was it even now shadowing them, moving soundlessly from cover to cover, like a tiger in the steel jungle ?", "label": "Contradiction", "premise": "The fact that I was hiding in the steel jungle, moving from one cover to the other, was like being scared, afraid and vulnerable", "explanation": "A tiger is a loud and ferocious animal, and its not scared, afraid and vulnerable."} 9 | {"id": "5", "hypothesis": "I attacked the problem as soon as I got out of bed.", "label": "Entailment", "premise": "I tried solving the problem as soon as I got out of bed.", "explanation": "Attacked is a metaphor meaning to try to solve something with determination as soon as possible."} 10 | {"id": "5", "hypothesis": "I attacked the problem as soon as I got out of bed.", "label": "Contradiction", "premise": "I put off doing my work until I had no choice left but to do it.", "explanation": "As soon as typically connotes immediacy while put off connotes delaying or procrastinating."} 11 | {"id": "6", "hypothesis": "Trembling for ire, and stricken pale for rage.", "label": "Entailment", "premise": "I can't stop shaking from the anger, and also I seem to be pale because of the rage.", "explanation": "The phrase trembling for ire means that the speaker is so angry that they are shaking while stricken with paleness."} 12 | {"id": "6", "hypothesis": "Trembling for ire, and stricken pale for rage.", "label": "Contradiction", "premise": "Trembling for ire, and disregard pale for rage.", "explanation": "One sentence compares the man's feelings to stricken with paleness whereas the other sentence contradicts this by saying that the man is actually disregarding it."} 13 | {"id": "7", "hypothesis": "Marcus enjoyed the book.", "label": "Entailment", "premise": "Marcus spent the whole day reading this book and was happy about it.", "explanation": "Marcus enjoyed the book means that he was happy about spending the whole day reading it."} 14 | {"id": "7", "hypothesis": "Marcus enjoyed the book.", "label": "Contradiction", "premise": "Marcus hated the book.", "explanation": "To enjoy something means to take pleasure in it, while to hate something means to despise or detest it."} 15 | {"id": "8", "hypothesis": "My love is a summer day.", "label": "Entailment", "premise": "My feelings of love make me very joyous just like a summer day.", "explanation": "A summer day is a time when the weather is warm and sunny and people are generally happy."} 16 | {"id": "8", "hypothesis": "My love is a summer day.", "label": "Contradiction", "premise": "My love is a long winter night.", "explanation": "The two statements are contradictory because a summer day is typically associated with warmth and happiness, while a long winter night is usually cold and dark."} 17 | {"id": "9", "hypothesis": "He was interwoven in a laughing fit.", "label": "Entailment", "premise": "He was involved in a spree for laughing so hard.", "explanation": "Laughing fits are often associated with a lack of control, while laughing very hard."} 18 | {"id": "9", "hypothesis": "He was interwoven in a laughing fit.", "label": "Contradiction", "premise": "He was uninvolved in a laughing fit.", "explanation": "To be interwoven in something means to be deeply or intimately connected to it, while to be uninvolved in something means to have no connection to it."} 19 | {"id": "10", "hypothesis": "I was drowned in work.", "label": "Entailment", "premise": "I was overwhelmed by work.", "explanation": "Drowned in work means that there was so much work that it was impossible to do it all."} 20 | {"id": "10", "hypothesis": "I was drowned in work.", "label": "Contradiction", "premise": "I was rescued from work.", "explanation": "Drowning is usually used to present a scenario of overwhelmed or submerged by something while being rescued connotes the opposite of being overwhelmed, which is being saved or relieved."} 21 | {"id": "11", "hypothesis": "I never won an argument with him.", "label": "Entailment", "premise": "He never gives up in an argument.", "explanation": "Not winning an argument means that you stop trying to argue your point and concede defeat."} 22 | {"id": "11", "hypothesis": "I never won an argument with him.", "label": "Contradiction", "premise": "I always win my arguments with him.", "explanation": "You can't always win your arguments with him if you've never won an argument with him."} 23 | {"id": "12", "hypothesis": "Only made the fire burn brighter ,", "label": "Entailment", "premise": "Only caused the fire to burn brighter ,", "explanation": "Made the fire burn brighter means that it caused the fire to become brighter."} 24 | {"id": "12", "hypothesis": "Only made he fire burn brighter ,", "label": "Contradiction", "premise": "Only stopped the fire to burn brighter .", "explanation": "Stopped the fire burn brighter means that it caused the fire to become weaker instead of brighter."} 25 | {"id": "13", "hypothesis": "All of this stirred an uncontrollable excitement in her.", "label": "Entailment", "premise": "The knowledge of this all put her in a very high state of emotions.", "explanation": "To be in a high state of emotions means to be very excited and in a state of uncontrollable excitements."} 26 | {"id": "13", "hypothesis": "All of this stirred an uncontrollable excitement in her.", "label": "Contradiction", "premise": "All of this stirred a dullness for her.", "explanation": "The two words uncontrollable excitement and dullness are opposite in meaning, so they contradict each other."} 27 | {"id": "14", "hypothesis": "The stains on the carpet have survived every name change .", "label": "Entailment", "premise": "Stains on the carpet managed to survive several name changes and still persist.", "explanation": "The stains on the carpet have been there for a long time and have survived every name change."} 28 | {"id": "14", "hypothesis": "The stains on the carpet have survived every name change .", "label": "Contradiction", "premise": "The stains on the carpet have succumbed to every name change.", "explanation": "To survive something means to continue to exist or live through it, while to succumb to something means to give in or yield to it."} 29 | {"id": "15", "hypothesis": "He is the apple of my eye.", "label": "Entailment", "premise": "I hold him dear and love him very much.", "explanation": "The apple of my eye means he is the person I love the most and hold dear."} 30 | {"id": "15", "hypothesis": "He is the apple of my eye.", "label": "Contradiction", "premise": "I strongly despise him.", "explanation": "The phrase apple of my eye is used to describe someone who is cherished and loved, where, in contrast, the phrase I strongly despise him indicates that the speaker has a deep hatred for the person in question."} 31 | {"id": "16", "hypothesis": "The dancer glided across the stage.", "label": "Entailment", "premise": "The dancer moved so fast that it seemed like she was across the stage.", "explanation": "The dancer moved so fast that it seemed like she was across the stage, here glided means to move smoothly and effortlessly across something."} 32 | {"id": "16", "hypothesis": "The dancer glided across the stage.", "label": "Contradiction", "premise": "The dancer paused at the middle of the stage.", "explanation": "To glide across something means to move smoothly and effortlessly, while to pause at something means to come to a stop."} 33 | {"id": "17", "hypothesis": "A weather vane crowns the building.", "label": "Entailment", "premise": "Weather vane tops off this magnificent building.", "explanation": "Crowns the building means it is sitting on the top of the building."} 34 | {"id": "17", "hypothesis": "A weather vane crowns the building.", "label": "Contradiction", "premise": "A weather vane mars the building.", "explanation": " A weather vane is a device that is used to show the direction of the wind, while something that mars a building would be considered an eyesore or blemish."} 35 | {"id": "18", "hypothesis": "When it rains in Tokyo the city becomes a sea of umbrellas.", "label": "Entailment", "premise": "When it rains in Tokyo, umbrellas are everywhere.", "explanation": "A sea of umbrellas means that there are so many umbrellas that they seem to fill up the entire city."} 36 | {"id": "18", "hypothesis": "When it rains in Tokyo the city becomes a sea of umbrellas.", "label": "Contradiction", "premise": "When it rains in Tokyo you won't find umbrellas in it.", "explanation": "Tokyo is a city with a lot of people in it and when it rains, everyone opens their umbrellas, exactly contradictory to the other statement ."} 37 | {"id": "19", "hypothesis": "She was lost in her thoughts and memories.", "label": "Entailment", "premise": "She was distracted by her thoughts and memories.", "explanation": "To be lost in one's thoughts and memories means to be thinking about the past and feel distracted about how things might have been different."} 38 | {"id": "19", "hypothesis": "She was lost in her thoughts and memories.", "label": "Contradiction", "premise": "She was attentive and present to the conversation.", "explanation": "To be lost in something means to be distracted or absent-minded, while to be attentive and present means to be focused and engaged."} 39 | {"id": "20", "hypothesis": "Their faces were blanketed with sadness.", "label": "Entailment", "premise": "Their faces were covered with sadness.", "explanation": "The phrase blanketed with sadness is a metaphor meaning that the person's faces were completely covered with sadness, like a blanket would cover a bed."} 40 | {"id": "20", "hypothesis": "Their faces were blanketed with sadness.", "label": "Contradiction", "premise": "Their faces lost all the sadness and sorrow.", "explanation": "To be blanketed with something means to be covered or hidden by it, while to lose something means to no longer have it."} 41 | {"id": "21", "hypothesis": "The play is going to pummel a few people.", "label": "Entailment", "premise": "A few people will be affected by the play.", "explanation": "Pummel means to hit something hard."} 42 | {"id": "21", "hypothesis": "The play is going to pummel a few people.", "label": "Contradiction", "premise": "The drama is going to make people laugh at the end.", "explanation": "To pummel something means to hit it repeatedly or violently, while to make people laugh is the opposite of that."} 43 | {"id": "22", "hypothesis": "His condescending attitude made my blood boil.", "label": "Entailment", "premise": "His snobbishness pissed me off.", "explanation": "condescending attitude means to showoff snobbishness that can make the speaker very angry."} 44 | {"id": "22", "hypothesis": "His condescending attitude made my blood boil.", "label": "Contradiction", "premise": "His kind demeanor melted my heart.", "explanation": "Condescending means behaving in a way that shows a lack of respect for others, while kind means having a caring and friendly nature."} 45 | {"id": "23", "hypothesis": "We could still be looking for it when they blow us out of the water.", "label": "Entailment", "premise": "We could still be looking for it when they surpassed us . ’", "explanation": "‘To blow someone out of the water’ is an idiom meaning to surpass someone very easily or soundly."} 46 | {"id": "23", "hypothesis": "We could still be looking for it when they blow us out of the water.", "label": "Contradiction", "premise": "We could still be looking for it because we beat them soundly.", "explanation": "To blow someone out of the water means to defeat them decisively, which is contradicted in the other statement ."} 47 | {"id": "24", "hypothesis": "An ancient anger exploded in his heart.", "label": "Entailment", "premise": "He was overwhelmed by an anger he had suppressed for a long time and thus he finally snapped when someone made him so mad.", "explanation": "This metaphor is explaining how the person was feeling a deep anger that they had been suppressing for a very long time (i.e., since an ancient time) and it finally came out."} 48 | {"id": "24", "hypothesis": "An ancient anger exploded in his heart.", "label": "Contradiction", "premise": "An ancient anger dissipated in his heart.", "explanation": "The two situations - anger exploding and anger dissipating - mentioned in the two sentences are opposite of each other, so they would be considered a contradiction."} 49 | {"id": "25", "hypothesis": "The critics hyped this broadway production.", "label": "Entailment", "premise": "The critics have pampered this mediocre production with undeserved praise.", "explanation": "The critics have given this production more attention than it deserves."} 50 | {"id": "25", "hypothesis": "The critics hyped this broadway production.", "label": "Contradiction", "premise": "The critics laughed off this broadway production, saying it will be a total flop.", "explanation": "To hype something means to promote it or build it up, while to laugh something off means to dismiss it or downplay it."} 51 | {"id": "26", "hypothesis": "Josh played Bach.", "label": "Entailment", "premise": "Josh played a sonata of the composer Bach.", "explanation": "Playing Bach means playing a music piece, such as a sonata wrote by Bach."} 52 | {"id": "26", "hypothesis": "Josh played Bach.", "label": "Contradiction", "premise": "Josh is part of a techno band.", "explanation": "Bach is a classical composer while techno is a type of electronic dance music."} 53 | {"id": "27", "hypothesis": "The crowd was a roaring river.", "label": "Entailment", "premise": "The huge crowd was making so much noise I just couldn't believe it.", "explanation": "A roaring river is a huge body of water that makes a lot of noise similar to the noise a huge crowd can make."} 54 | {"id": "27", "hypothesis": "The crowd was a roaring river.", "label": "Contradiction", "premise": "The crowd was a quiet pond.", "explanation": "A roaring river is full of noise and movement, while a quiet pond is still and peaceful, thus the statements are contradictory."} 55 | {"id": "28", "hypothesis": "He pleaded for her forgiveness but Janet was a statue.", "label": "Entailment", "premise": "He made a heartfelt appeal to her but Janet was stone-faced and didn't give any sign of emotion.", "explanation": "Being statue means showing no emotion, i.e., Janet was not moved by his appeal and remained emotionless."} 56 | {"id": "28", "hypothesis": "He pleaded for her forgiveness but Janet was a statue.", "label": "Contradiction", "premise": "He pleaded for her forgiveness and Janet smiled.", "explanation": "In the one statement he pleaded for her forgiveness, but Janet was unresponsive to his pleas, or uninterested in hearing them, whereas, the other statement informs that Janet responded by smiling."} 57 | {"id": "29", "hypothesis": "Nobody cares about the roots of modern civilization.", "label": "Entailment", "premise": "Looking at history and realizing most people are ignorant to the origin of modern civilization.", "explanation": "The roots of modern civilization are the origins of the things that make up our modern civilization."} 58 | {"id": "29", "hypothesis": "Nobody cares about the roots of modern civilization.", "label": "Contradiction", "premise": "Everybody cares about the roots of modern civilization.", "explanation": "One statement suggests that people do care about the roots of modern civilization, which would contradict the other statement of the statement Nobody cares about the roots of modern civilization."} 59 | {"id": "30", "hypothesis": "Now he melted them to pity.", "label": "Entailment", "premise": "Now he humbled them to pity .", "explanation": "Melted means to make them feel emotions like pity."} 60 | {"id": "30", "hypothesis": "Now he melted them to pity.", "label": "Contradiction", "premise": "When he hardened their hearts against mercy.", "explanation": "To melt something means to make it soft or pliable, while to harden something means to make it tough or unyielding."} 61 | {"id": "31", "hypothesis": "Till the darkness falls around us ,", "label": "Entailment", "premise": "I long for the darkness to fall around me.", "explanation": "To long for the darkness to fall around us means we want the darkness to engulf us."} 62 | {"id": "31", "hypothesis": "Till the darkness falls around us ,", "label": "Contradiction", "premise": "The darkness rose from the ground .", "explanation": "Darkness typically falls from the sky as the sun sets, while the ground is where darkness would rise from if the sun was rising."} 63 | {"id": "32", "hypothesis": "She had a bright idea.", "label": "Entailment", "premise": "She came up with a brilliant idea.", "explanation": "A bright idea is something that is creative and clever."} 64 | {"id": "32", "hypothesis": "She had a bright idea.", "label": "Contradiction", "premise": "She was in a dark mood.", "explanation": "A person cannot be in a dark mood and have a bright idea at the same time."} 65 | {"id": "33", "hypothesis": "The computer in the classroom was a dinosaur.", "label": "Entailment", "premise": "The computer in the classroom is a very old model, hence the slow speed and all is to be expected.", "explanation": "Dinosaurs are very old, and they are known for being slow, thus it makes sense that an old, slow computer would be described as a dinosaur."} 66 | {"id": "33", "hypothesis": "The computer in the classroom was a dinosaur.", "label": "Contradiction", "premise": "The computer in the classroom was a revolutionary.", "explanation": "In the first statememt, the computer in the classroom refers to an old, outdated computer whereas the computer in the other statement refers to a new, innovative computer, hence they are contradictory."} 67 | {"id": "34", "hypothesis": "No,he said helpfully, then bestowed a patronising smile on me .", "label": "Entailment", "premise": "He blatantly said no and gave me a patronising smile .", "explanation": "To bestow a smile patronisingly is to give a smile that shows you think you are better than the other person."} 68 | {"id": "34", "hypothesis": "No, he said helpfully, then bestowed a patronising smile on me .", "label": "Contradiction", "premise": "Yes,' he said wickedly . Then he looked serious .", "explanation": "To say something with a patronising smile mean it is in a way that is helpful or useful, while to say something wickedly would be to say it in a way that is evil or malicious."} 69 | {"id": "35", "hypothesis": "He absorbed the knowledge or beliefs of his tribe.", "label": "Entailment", "premise": "He mentally assimilated the knowledge or beliefs of his tribe .", "explanation": "To absorb something is to take it in and make it part of yourself."} 70 | {"id": "35", "hypothesis": "He absorbed the knowledge or beliefs of his tribe.", "label": "Contradiction", "premise": "He utterly decimated his tribe's most deeply held beliefs.", "explanation": "Absorbed typically means to take in or take up something, while utterly decimated means to destroy completely."} 71 | {"id": "36", "hypothesis": "His friend was very tall and he looked like a chess piece beside him.", "label": "Entailment", "premise": "That huge guy beside him made him look so puny.", "explanation": "A chess piece is much smaller than the board it is on, so when one person is standing next to another who is much taller, they can look like a chess piece."} 72 | {"id": "36", "hypothesis": "His friend was very tall and he looked like a chess piece beside him.", "label": "Contradiction", "premise": "His friend was very short and he looked like a basketball player next to him.", "explanation": "The friend was very tall and the other person himself would look like a small chess piece next to the tall friend, which cannot be possible if the other statement is true which says he looks like a very tall person (basketball players are usually very tall) fromt the perspective of the friend."} 73 | {"id": "37", "hypothesis": "Ask table 12 whether he wants to swap.", "label": "Entailment", "premise": "Ask the person on table 12 whether he wants to exchange or not.", "explanation": "Asking someone to swap mean whether they want to exchange their sitting position."} 74 | {"id": "37", "hypothesis": "Ask table 12 whether he wants to swap.", "label": "Contradiction", "premise": "Ask customers at table 12 whether they want to keep the table.", "explanation": "Asking someone to swap mean whether they want to exchange their sitting position while keeping the table suggests the opposite."} 75 | {"id": "38", "hypothesis": "She captured all the men's hearts.", "label": "Entailment", "premise": "She owned all the men's hearts at that party.", "explanation": "When you capture someone's heart, it means you have taken control of their emotions."} 76 | {"id": "38", "hypothesis": "She captured all the men's hearts.", "label": "Contradiction", "premise": "She could not give a damn about the men in her life.", "explanation": "To capture someone's heart means to win their love or affection, while to not give a damn about someone means to not care about them at all."} 77 | {"id": "39", "hypothesis": "It dawned to him that she had betrayed him.", "label": "Entailment", "premise": "It occurs to him that she betrayed him.", "explanation": "It dawned on him means he realized or came to the understanding that she had betrayed him."} 78 | {"id": "39", "hypothesis": "It dawned to him that she had betrayed him.", "label": "Contradiction", "premise": "It passed him that she had betrayed him.", "explanation": "To dawn on someone means to become apparent or clear to them, while to pass someone means to go by without being noticed."} 79 | {"id": "40", "hypothesis": "She petitioned the witnesses for information.", "label": "Entailment", "premise": "She asked the witnesses for information.", "explanation": "Petitioning means to request or beg for something."} 80 | {"id": "40", "hypothesis": "She petitioned the witnesses for information.", "label": "Contradiction", "premise": "She was guarding the witnesses to protect information.", "explanation": "To petition something means to request it earnestly or formally, while to guard something means to protect it."} 81 | {"id": "41", "hypothesis": "Seurat culminated pointillism.", "label": "Entailment", "premise": "Seurat was the climax of pointillism .", "explanation": "Culminated means he was the highest point or the best of pointillism."} 82 | {"id": "41", "hypothesis": "Seurat culminated pointillism.", "label": "Contradiction", "premise": "Seurat was a pioneer of the pointillism Movement.", "explanation": "The word pioneer typically connotes someone who is first or early in a movement while culminated typically connotes someone who is at the end or late in a movement."} 83 | {"id": "42", "hypothesis": "Water cascaded all over the floor.", "label": "Entailment", "premise": "Water got all over the floor.", "explanation": "Cascaded means to fall or pour down in large amounts."} 84 | {"id": "42", "hypothesis": "Water cascaded all over the floor.", "label": "Contradiction", "premise": "Water backed up all over the floor.", "explanation": "To cascade means to flow in a constant stream, while to back up means to fill up or overflow."} 85 | {"id": "43", "hypothesis": "Now you can fade off to sleep.", "label": "Entailment", "premise": "You better go to sleep.", "explanation": "fade off to sleep means to go to sleep gradually."} 86 | {"id": "43", "hypothesis": "Now you can fade off to sleep.", "label": "Contradiction", "premise": "Now you can't sleep.", "explanation": "One statement is saying that you can sleep, while the other statement is contradicting the first."} 87 | {"id": "44", "hypothesis": "She arrived at the grocery store with an army of children.", "label": "Entailment", "premise": "She arrived at the grocery store with a massive group of children.", "explanation": "An army of children means a large group of children."} 88 | {"id": "44", "hypothesis": "She arrived at the grocery store with an army of children.", "label": "Contradiction", "premise": "She arrived at the grocery store with no children.", "explanation": "One sentence implies that she arrived at the store with a large number of children, while the other sentence implies that she arrived at the store alone."} 89 | {"id": "45", "hypothesis": "Rohan bought a Picasso.", "label": "Entailment", "premise": "Yeah, Rohan spent so much on a Picasso painting.", "explanation": "Rohan buying a Picasso means he bought a painting by Picasso."} 90 | {"id": "45", "hypothesis": "Rohan bought a Picasso.", "label": "Contradiction", "premise": "Rohan has a poor choice in terms of paintings.", "explanation": "Picasso is a famous and well-respected artist, while having a poor choice in paintings would suggest that Rohan does not have good taste in art, i.e., won't buy a painting of Picasso."} 91 | {"id": "46", "hypothesis": "Nor, if we are in civilised cast of mind, because we wish to become drunk .", "label": "Entailment", "premise": "Nor, if we are in civilised mindset, because we wish to become drunk .", "explanation": "Civilised cast of mind means that we are thinking in a way that is rational and civilized, i.e., we would not want to become drunk because that would be irrational and uncivilized."} 92 | {"id": "46", "hypothesis": "Nor, if we are in civilised cast of mind, because we wish to become drunk .", "label": "Contradiction", "premise": "Nor, if we are in a civilised but skeptical state of mind, because we wish to become drunk.", "explanation": "If we wish to become sober, it would suggest that we are behaving in a civilised manner, which is opposed in the other statement saying the mindset is skeptical."} 93 | {"id": "47", "hypothesis": "The republican convention was a horror show.", "label": "Entailment", "premise": "The process of how Republicans hold their convention was very distressting and primitive.", "explanation": "The phrase a horror show is often used to describe something that is very distressing or upsetting, so, in this case, the author is saying that the Republican convention was very distressing and primitive."} 94 | {"id": "47", "hypothesis": "The republican convention was a horror show.", "label": "Contradiction", "premise": "The republican convention was a joyous and peaceful affair.", "explanation": "A horror show is something that is full of terror or dread, while a joyous and peaceful affair is something that is happy and calm."} 95 | {"id": "48", "hypothesis": "Water seeped out of the can into the backpack.", "label": "Entailment", "premise": "Canned water leaked into the backpack.", "explanation": "Can refers to a container that holds liquid and is sealed, here, when water seeps out, it means the liquid has leaked."} 96 | {"id": "48", "hypothesis": "Water seeped out of the can into the backpack.", "label": "Contradiction", "premise": "Water accidentally entered into my backpack.", "explanation": "Seepage is a gradual process where water or another liquid seeps or leaks through something, while water entering accidentally would suggest that it was not a gradual process."} 97 | {"id": "49", "hypothesis": "When I met the boss he seemed bitter about the whole story.", "label": "Entailment", "premise": "He seemed quite annoyed after hearing about the full story.", "explanation": "Feeling bitter means the boss was likely annoyed after hearing about the story because it wasn't what he wanted to hear."} 98 | {"id": "49", "hypothesis": "When I met the boss he seemed bitter about the whole story.", "label": "Contradiction", "premise": "When I met the boss he seemed about to burst with happiness.", "explanation": "Bitter is the opposite of happy, so these two emotions would not be felt at the same time."} 99 | {"id": "50", "hypothesis": "Time flies.", "label": "Entailment", "premise": "Time seems to pass very quickly.", "explanation": "This metaphor means that time seems to move very quickly."} 100 | {"id": "50", "hypothesis": "Time flies.", "label": "Contradiction", "premise": "Time creeps by.", "explanation": "Time flies means that time passes by quickly, while time creeps by means that time passes by slowly."} 101 | {"id": "51", "hypothesis": "The sun glared down on us.", "label": "Entailment", "premise": "The sun reflecting a dazzling light to us.", "explanation": "Glaring is used to mean that the sun is reflecting a strong light."} 102 | {"id": "51", "hypothesis": "The sun glared down on us.", "label": "Contradiction", "premise": "The Sun was shy and hid behind a cloud.", "explanation": "The sun glaring down on us would suggest that it is shining brightly and harshly, while the sun being shy and hiding behind a cloud would suggest that it is not shining very brightly."} 103 | {"id": "52", "hypothesis": "The work would have been easier if I had an extra pair of hands.", "label": "Entailment", "premise": "Had I enough workers, I could have completed my work in half the time.", "explanation": "More workers would mean more hands to do the work, and the work would be completed faster."} 104 | {"id": "52", "hypothesis": "The work would have been easier if I had an extra pair of hands.", "label": "Contradiction", "premise": "The work would have been more easier if I could do it own my own", "explanation": "Having an extra pair of hands would typically make a task easier because there would be more people to help with it which is cotradictory to doing it on someone's own."} 105 | {"id": "53", "hypothesis": "He kept us in the dark about his plans.", "label": "Entailment", "premise": "He did not share any information about his plans.", "explanation": "In the dark - means not sharing the information or something."} 106 | {"id": "53", "hypothesis": "He kept us in the dark about his plans.", "label": "Contradiction", "premise": "He shared his plans with us in details.", "explanation": "The speaker says that the person kept them in the dark, it means that the person did not tell them what was happening or what was going to happen, whereas, if the speaker says that the person shared their plans with them in detail, it means that the person told them everything about what was going to happen."} 107 | {"id": "54", "hypothesis": "You are trying my patience.", "label": "Entailment", "premise": "You make me feel like losing my patience.", "explanation": "When someone says you are trying my patience, they mean that you are doing something that is making them angry and they may loose their patience."} 108 | {"id": "54", "hypothesis": "You are trying my patience.", "label": "Contradiction", "premise": "You're building my patience.", "explanation": "Trying someone's patience means to test their ability to tolerate or endure something, while building someone's patience would suggest the opposite of making them less tense."} 109 | {"id": "55", "hypothesis": "Raged and roared the wintry tempest ,", "label": "Entailment", "premise": "Raged and made a loud resounding noise the wintry tempest.", "explanation": "Roared means to make a loud noise"} 110 | {"id": "55", "hypothesis": "Raged and roared the wintry tempest ,", "label": "Contradiction", "premise": "The wintry tempest so softly and gently whimpered.", "explanation": "Roared means to make a loud noise while whimpering typically mean a feeble low sound that express fear."} 111 | {"id": "56", "hypothesis": "Andrew's emails were beating a dead horse.", "label": "Entailment", "premise": "It is not a success because none of Andrew's emails got replied.", "explanation": "Beating a dead horse means trying to do something that is no longer possible or useful, which is similar to Andrew's emails because he was not getting any replies."} 112 | {"id": "56", "hypothesis": "Andrew's emails were beating a dead horse.", "label": "Contradiction", "premise": "Andrew's emails were easy to read and had a positive outcome.", "explanation": "To beat a dead horse means to continue doing something even though it is no longer effective, such as Andrew's repetitive emils while emails that are easy to read and have a positive outcome would be the opposite of beating a dead horse."} 113 | {"id": "57", "hypothesis": "Now he stirred their souls to passion ,", "label": "Entailment", "premise": "Now he made their souls passionate ,", "explanation": "stirred means to move emotionally or passionately."} 114 | {"id": "57", "hypothesis": "Now he stirred their souls to passion ,", "label": "Contradiction", "premise": "Now he lowered their expectations,", "explanation": "To stir someone's soul to passion means to excite or inspire them, while to lower someone's expectations means to disappoint them."} 115 | {"id": "58", "hypothesis": "Fire had devoured our home.", "label": "Entailment", "premise": "Fire had taken the whole house.", "explanation": "Devoured means to consume or destroy something completely."} 116 | {"id": "58", "hypothesis": "Fire had devoured our home.", "label": "Contradiction", "premise": "Fire had warmed our home.", "explanation": "The word devoured means something that is destroyed or consumed, while warmed typically connotes something that is made more comfortable or cozy."} 117 | {"id": "59", "hypothesis": "She did not realize that opportunity was waiting at her door.", "label": "Entailment", "premise": "She totally missed an opportunity to present itself at her doorstep.", "explanation": "When we say that opportunity was waiting at her door, we mean that it was there for her to take advantage of, but she did not realize it."} 118 | {"id": "59", "hypothesis": "She did not realize that opportunity was waiting at her door.", "label": "Contradiction", "premise": "She left at opportunity just sitting there and didn't bother.", "explanation": "An opportunity typically doesn't wait at someone's door, but rather it is something that someone has to seek out or take advantage of."} 119 | {"id": "60", "hypothesis": "The force charged the village at night.", "label": "Entailment", "premise": "The force attacked the poor village during the night.", "explanation": "Charging is an aggressive action of attacking, often done by an army, and it can be done at night."} 120 | {"id": "60", "hypothesis": "The force charged the village at night.", "label": "Contradiction", "premise": "The village was protected by force.", "explanation": "When the force charged the village, it means they were attacking it, whereas, the other statement contradicts saying, the force were actually the ones protecting the village."} 121 | {"id": "61", "hypothesis": "The candidate rammed into his opponent mercilessly.", "label": "Entailment", "premise": "The candidate tried usinsg harsh language against his opponent in order to defeat him.", "explanation": "Rammed into his opponent mercilessly means the candidate attacked his opponent without mercy."} 122 | {"id": "61", "hypothesis": "The candidate rammed into his opponent mercilessly.", "label": "Contradiction", "premise": "The candidate protected his opponent.", "explanation": "To ram into someone means to attack them forcefully, while to protect someone means to defend or shield them."} 123 | {"id": "62", "hypothesis": "He weathered the costs for the accident.", "label": "Entailment", "premise": "He has to bear the costs for the accident.", "explanation": "To weather something is to endure it or suffer through it."} 124 | {"id": "62", "hypothesis": "He weathered the costs for the accident.", "label": "Contradiction", "premise": "He got away with not paying all the costs for the accident.", "explanation": "To weather something means to endure or withstand it, while to get away with something means to escape or avoid it."} 125 | {"id": "63", "hypothesis": "The car hollered as the key was roughly turned in its ignition.", "label": "Entailment", "premise": "The car honked as the key was roughly turned in its ignition.", "explanation": "The word holler is a metaphor, and it is used to describe a louder, more aggressive sound."} 126 | {"id": "63", "hypothesis": "The car hollered as the key was roughly turned in its ignition.", "label": "Contradiction", "premise": "The car broke down and did not make any noise because the key was roughly turned in its ignition.", "explanation": "To holler means to make a loud noise, while to break down means to stop working."} 127 | {"id": "64", "hypothesis": "The night sky flared with the massive bombardment.", "label": "Entailment", "premise": "Such devastating bombardment from the night sky and it was massive .", "explanation": "The night sky flared means there was a lot of activity and movement in the sky."} 128 | {"id": "64", "hypothesis": "The night sky flared with the massive bombardment.", "label": "Contradiction", "premise": "The night sky was devoid of the usual huge fireworks.", "explanation": "Flare is used here to mean something that is bright and explosive, whereas, devoid means empty or void of something."} 129 | {"id": "65", "hypothesis": "The children twinkled with joy.", "label": "Entailment", "premise": "The children were shining with joy.", "explanation": "To twinkle with joy means to be filled with happiness and shining."} 130 | {"id": "65", "hypothesis": "The children twinkled with joy.", "label": "Contradiction", "premise": "The children were bored and hence sulked.", "explanation": "To twinkle with joy means to be filled with happiness and excitement, while to sulk means to be sulky or ill-tempered."} 131 | {"id": "66", "hypothesis": "Stock prices are drifting higher.", "label": "Entailment", "premise": "The entire stock market is going higher.", "explanation": "Stock prices are drifting higher means an upward gradual movement."} 132 | {"id": "66", "hypothesis": "Stock prices are drifting higher.", "label": "Contradiction", "premise": "Stocks are going down.", "explanation": "The phrase drifting higher ,means a gradual movement to the top while going down connotes a sharp decrease."} 133 | {"id": "67", "hypothesis": "We still ignore how the seed of life was planted on our planet.", "label": "Entailment", "premise": "We still ignore how life was started on our planet.", "explanation": "The seed of life means the beginning of life."} 134 | {"id": "67", "hypothesis": "We still ignore how the seed of life was planted on our planet.", "label": "Contradiction", "premise": "We still ignore how the idea of death was planted on our planet.", "explanation": "The seed of life means how life was started on our planet where as the idea of death is its exact opposite."} 135 | {"id": "68", "hypothesis": "The needle bit his skin.", "label": "Entailment", "premise": "The needle pricked his skin.", "explanation": "The needle bit his skin means that the needle punctured his skin."} 136 | {"id": "68", "hypothesis": "The needle bit his skin.", "label": "Contradiction", "premise": "The needle jumped away to avoid the protagonist's skin.", "explanation": "When a needle bites someone, it means that it punctures their skin, but if the needle were to jump away, it would not be able to puncture the skin."} 137 | {"id": "69", "hypothesis": "You must adhere to the rules.", "label": "Entailment", "premise": "You must abide by the rules.", "explanation": "Adhere to the rules means to follow the rules."} 138 | {"id": "69", "hypothesis": "You must adhere to the rules.", "label": "Contradiction", "premise": "You must break the rules.", "explanation": "adhere to the rules mean following or complying with the rules while break typically connotes going against or disregarding the rules."} 139 | {"id": "70", "hypothesis": "Came the Spring with all its splendor ,", "label": "Entailment", "premise": "Came spring, with all its sweetness in form of bright colors.", "explanation": "Splendor is used here to mean that the Spring is a time when we see all the magnificent features or qualities."} 140 | {"id": "70", "hypothesis": "Came the Spring with all its splendor ,", "label": "Contradiction", "premise": "Came the Spring with all its blunder,", "explanation": "Splendor is a word that is often used to describe something that is beautiful or magnificent, while blunder is a word that is used to describe something that is clumsy or inept."} 141 | {"id": "71", "hypothesis": "I want to frolic in sunshine once again.", "label": "Entailment", "premise": "I want to feel happy in the sunshine again.", "explanation": "Frolic means to play or dance around in a happy and carefree manner."} 142 | {"id": "71", "hypothesis": "I want to frolic in sunshine once again.", "label": "Contradiction", "premise": "I want to be able to hide in sunshine again.", "explanation": "To frolic in something means to play or have fun in it, while to hide in something means to take shelter or refuge in it."} 143 | {"id": "72", "hypothesis": "The crowd flowed out of the stadium.", "label": "Entailment", "premise": "The endless wave of the crowd rushing out of the stadium.", "explanation": "The metaphor is saying that the crowd flowed out of the stadium, which means that there was a lot of people leaving the stadium all at once."} 144 | {"id": "72", "hypothesis": "The crowd flowed out of the stadium.", "label": "Contradiction", "premise": "The crowd marched in to the stadium.", "explanation": "Flowed out typically means a smooth and gentle movement of the crowd going out of the stadium while marched in is a more forceful and aggressive movement of the crowd going in the stadium."} 145 | {"id": "73", "hypothesis": "By that time my friend was broken so he had to sell his house.", "label": "Entailment", "premise": "That was a situation where my friend had to sell his house because he had no money.", "explanation": "Broken in this context means that the person is financially ruined."} 146 | {"id": "73", "hypothesis": "By that time my friend was broken so he had to sell his house.", "label": "Contradiction", "premise": "By that time my friend was wealthy and he bought another house.", "explanation": "The speaker's friend was not wealthy at the time mentioned, but rather was in a difficult financial situation and had to sell their house, which contradicts the claim that the friend was wealthy and bought another house."} 147 | {"id": "74", "hypothesis": "Sniffles, coughs, and colds gurgle in their breasts.", "label": "Entailment", "premise": "Every time someone has a cold, they develop a cough and a sniffle at the same time.", "explanation": "The phrase gurgle in their breasts is a metaphor meaning that the person has a lot of mucus in their chest."} 148 | {"id": "74", "hypothesis": "Sniffles, coughs, and colds gurgle in their breasts.", "label": "Contradiction", "premise": "Sniffles, coughs, and colds reduced in their breasts.", "explanation": "The phrase gurgle in their breasts is a metaphor meaning that the person has a lot of mucus in their chest which contradicts that Sniffles, coughs, and colds have rediced in their ches."} 149 | {"id": "75", "hypothesis": "You bring back my youth of passion ,", "label": "Entailment", "premise": "You restore my youth of passion ,", "explanation": "bring back means to make something happen again."} 150 | {"id": "75", "hypothesis": "You bring back my youth of passion ,", "label": "Contradiction", "premise": "You make me feel old and tired.", "explanation": "When someone brings back your youth, it means they make you feel young again, while getting old is the opposite of feeling young."} 151 | {"id": "76", "hypothesis": "The show was characterised by dark humor and satire.", "label": "Entailment", "premise": "The show was known to joke about matters considered taboo and for its use of satire.", "explanation": "The word characterize is used here to showcase that the show was identified by its use of humor that pokes fun at serious topics."} 152 | {"id": "76", "hypothesis": "The show was characterised by dark humor and satire.", "label": "Contradiction", "premise": "The show did not have any specific bit of dark humor or satire.", "explanation": "In one statement it says the show was characterized by the use of dark humor whereas the other statement says the show did not have any specific use of dark humor or satire."} 153 | {"id": "77", "hypothesis": "The sun was a toddler insistently refusing to go to bed: It was past eight thirty and still light.", "label": "Entailment", "premise": "The sun was not setting, it was still light outside even if it was very late.", "explanation": "Toddlers do not like to go to bed when it is still light outside even if it past their bed time."} 154 | {"id": "77", "hypothesis": "The sun was a toddler insistently refusing to go to bed: It was past eight thirty and still light.", "label": "Contradiction", "premise": "The sun was a toddler happily tucking itself into bed : It was only three in the afternoon.", "explanation": "One statement is comparing the sun to a toddler who is staying up past their bedtime whereas the other statement compares the sun to a toddler who is happily going to bed, which is the opposite of what the one statement is saying."} 155 | {"id": "78", "hypothesis": "I was home watching the days slip by from my window.", "label": "Entailment", "premise": "I was home doing nothing while days passed.", "explanation": "Days slip by means the days passed by quickly."} 156 | {"id": "78", "hypothesis": "I was home watching the days slip by from my window.", "label": "Contradiction", "premise": "I was home watching the days race by from my window.", "explanation": "The original sentence suggests that the speaker was not doing anything active, but simply observing time passing, that contradicts the other statement that suggests that the speaker was watching time fly by, implying that they were somehow active or engaged in what they were doing."} 157 | {"id": "79", "hypothesis": "The distinction between the two theories blurred.", "label": "Entailment", "premise": "It's hard to differentiate the two theories given they are so alike.", "explanation": "Blurred means to break down or disappear, i.e., in this context, it means that the difference between the two theories is no longer clear and vanished."} 158 | {"id": "79", "hypothesis": "The distinction between the two theories blurred.", "label": "Contradiction", "premise": "The distinction between the two theories grew evident.", "explanation": "The word blurred typically connotes something that is not clear or evident, while grew evident means that the distinction became more clear."} 159 | {"id": "80", "hypothesis": "She was besieged by so many problems that she got discouraged.", "label": "Entailment", "premise": "She got overwhelmed by a bunch of problems and became discouraged.", "explanation": "Besieged means to be surrounded by something, in this case problems, so when someone is besieged, it can be overwhelming and discouraging."} 160 | {"id": "80", "hypothesis": "She was besieged by so many problems that she got discouraged.", "label": "Contradiction", "premise": "She was unbothered by the problems.", "explanation": "When a person is besieged by problems, it typically means that they are overwhelmed and discouraged, whereas, on the contrary, the word unbothered means she was unaffected by the problems."} 161 | {"id": "81", "hypothesis": "Inflate the currency.", "label": "Entailment", "premise": "Increase the quantity of the currency.", "explanation": "Inflation is when there is too much money in circulation and the value of the currency decreases."} 162 | {"id": "81", "hypothesis": "Inflate the currency.", "label": "Contradiction", "premise": "Deflate the currency.", "explanation": "Inflating the currency means to increase the money supply in circulation which would lead to higher prices, while deflating the currency means to decrease the money supply in circulation which would lead to lower prices."} 163 | {"id": "82", "hypothesis": "The dancer breezed across the stage.", "label": "Entailment", "premise": "The dancer is trying to run around the stage as fast as possible.", "explanation": "Breezed across the stage means the dancer is moving quickly across the stage."} 164 | {"id": "82", "hypothesis": "The dancer breezed across the stage.", "label": "Contradiction", "premise": "The dancer paused in the middle of the stage.", "explanation": "To breeze across something means to move quickly and effortlessly, while to pause in the middle of something suggests that there is a obstruction or a hesitation."} 165 | {"id": "83", "hypothesis": "For the Summer-time is coming ,", "label": "Entailment", "premise": "For the Summer-time will soon begin ,", "explanation": "coming means it is getting closer ."} 166 | {"id": "83", "hypothesis": "For the Summer-time is coming ,", "label": "Contradiction", "premise": "For this summer is coming to an end.", "explanation": "Summertime is typically seen as a time of warmth and sunshine, while the end of summer is seen as a time when the weather starts to get cooler."} 167 | {"id": "84", "hypothesis": "This office is buzzing with activity.", "label": "Entailment", "premise": "This office is as lively as a bustling beehive.", "explanation": "A beehive is a place where bees live and buzz."} 168 | {"id": "84", "hypothesis": "This office is buzzing with activity.", "label": "Contradiction", "premise": "This office is desolate due to the abandonment of work.", "explanation": "The word buzzing typically connotes a lot of activity and movement while desolate means empty or void of people."} 169 | {"id": "85", "hypothesis": "In the depth of the night I sink in the bed.", "label": "Entailment", "premise": "In the depth of the night I lie in the bed.", "explanation": "Sink in the bed means to become engulfed by the bed, to become lost in its depths."} 170 | {"id": "85", "hypothesis": "In the depth of the night I sink in the bed.", "label": "Contradiction", "premise": "At midnight I wake up from my bed.", "explanation": "To sink in something means to go down or disappear into it, while to wake up from something means to become conscious or aware of it."} 171 | {"id": "86", "hypothesis": "She said he was very sweet and he smiled brightly.", "label": "Entailment", "premise": "She said he was very nice and he beamed a smile at her.", "explanation": "Sweet is used to mean the person is very nice."} 172 | {"id": "86", "hypothesis": "She said he was very sweet and he smiled brightly.", "label": "Contradiction", "premise": "She said he was very sweet and he frowned.", "explanation": "One statement says that the person was sweet and smiled, while the second contradicts it by saying that the person was sweet and frowned."} 173 | {"id": "87", "hypothesis": "I toiled through the work.", "label": "Entailment", "premise": "I struggled through the work.", "explanation": "Toil means to work hard, and often refers to physical labor, so, here, in this context, it emphasizes the speaker's feeling of exhaustion and hopelessness."} 174 | {"id": "87", "hypothesis": "I toiled through the work.", "label": "Contradiction", "premise": "I ace through the work.", "explanation": "To toil through something means to work hard and laboriously, while to ace through something means to do it easily or effortlessly."} 175 | {"id": "88", "hypothesis": "Joseph has the heart of a lion.", "label": "Entailment", "premise": "Joseph is such a couragious guy.", "explanation": "Lions are known for their courage."} 176 | {"id": "88", "hypothesis": "Joseph has the heart of a lion.", "label": "Contradiction", "premise": "Joseph has the calm demeanor of a lamb.", "explanation": "A lion is known for being a fierce and powerful animal, while a lamb is known for being gentle and timid."} 177 | {"id": "89", "hypothesis": "The needle nipped his skin.", "label": "Entailment", "premise": "His skin was poked by the needle.", "explanation": "Nipped means to poke or stab."} 178 | {"id": "89", "hypothesis": "The needle nipped his skin.", "label": "Contradiction", "premise": "The needle avoided his skin.", "explanation": "To nip something means to cut or bite it, while to avoid something means to stay away from it or keep it at a distance."} 179 | {"id": "90", "hypothesis": "William loved cigarettes.", "label": "Entailment", "premise": "William always got hooked on cigarettes.", "explanation": "William always got hooked on cigarettes means that he was always addicted to cigarettes and loved smoking them."} 180 | {"id": "90", "hypothesis": "William loved cigarettes.", "label": "Contradiction", "premise": "William hated cigarettes.", "explanation": "To love something means to have strong feelings of affection towards it, while to hate something means to feel intense or passionate dislike for it."} 181 | {"id": "91", "hypothesis": "Fame clings to the heroes.", "label": "Entailment", "premise": "Fame and fortune are for real heroes.", "explanation": "Clinging means to attach or stick to something, so, here, the fame is attaching itself to the heroes."} 182 | {"id": "91", "hypothesis": "Fame clings to the heroes.", "label": "Contradiction", "premise": "Fame avoids the heroes.", "explanation": "Fame is often seen as something that clings to or follows someone who is successful or important, while avoiding someone would suggest that fame is not associated with them."} 183 | {"id": "92", "hypothesis": "The car bleated as the key was roughly turned in its ignition.", "label": "Entailment", "premise": "The car made noises as I roughly turned the key in its ignition.", "explanation": "The car is bleating like a sheep, which is known for making noises."} 184 | {"id": "92", "hypothesis": "The car bleated as the key was roughly turned in its ignition.", "label": "Contradiction", "premise": "The car broke down and made no sound because the person turned the key in the ignition very roughly.", "explanation": "To bleat means to make a sound like a sheep, while to break down means to stop working."} 185 | {"id": "93", "hypothesis": "The wind howled; the waves dashed their bucklers together - we were in the jaws of death.", "label": "Entailment", "premise": "The wind roared; the waves collided - we were in great danger.", "explanation": "The metaphor here is comparing the storm to a dangerous animal, with the waves being its teeth."} 186 | {"id": "93", "hypothesis": "The wind howled; the waves dashed their bucklers together - we were in the jaws of death.", "label": "Contradiction", "premise": "The wind whistled; the waves swished through their pearly gates - we were in the womb of an angel.", "explanation": "One sentence suggests that we are in a dangerous situation while the other sentence suggests that we are in a safe and protected place."} 187 | {"id": "94", "hypothesis": "Fill the child with pride.", "label": "Entailment", "premise": "Make the child feel proud.", "explanation": "When you fill something, it means to put something into it, so in this sentence, the author is saying that the child will feel pride, or have a lot of pride."} 188 | {"id": "94", "hypothesis": "Fill the child with pride.", "label": "Contradiction", "premise": "Empty the child with pride.", "explanation": "When you fill something, it means to put something into it, whereas emptying means unfilling it."} 189 | {"id": "95", "hypothesis": "You were dying to believe in what you heard.", "label": "Entailment", "premise": "You liked what you heard so much you just wanted to believe it.", "explanation": "To be dying to do something means you really want to do it."} 190 | {"id": "95", "hypothesis": "You were dying to believe in what you heard.", "label": "Contradiction", "premise": "You couldn't have wanted to believe in what you heard.", "explanation": "To be dying to do something means to really want to do it, while to not want to do something would suggest that you are indifferent or uninterested in it."} 191 | {"id": "96", "hypothesis": "I bet you have not done your homework.", "label": "Entailment", "premise": "It is a fair assumption to say that you didn't do your homework..", "explanation": "The word bet is used here to mean the speaker has a fair assumption of the situation."} 192 | {"id": "96", "hypothesis": "I bet you have not done your homework.", "label": "Contradiction", "premise": "I doubt you didn't do your homework.", "explanation": "One sentence suggesting that the speaker has a strong assumption that the other person has not done their homework whereas the other sentence contradicts this by saying that now the speaker is doubtful."} 193 | {"id": "97", "hypothesis": "And then added, speaking slowly ,", "label": "Entailment", "premise": "And then said additionally, speaking slowly ,", "explanation": "added means to say something in addition to what has already been said ."} 194 | {"id": "97", "hypothesis": "And then added, speaking slowly ,", "label": "Contradiction", "premise": "And then to the contrary, speaking slowly.", "explanation": "added means to say something in addition whereas to the contrary means its opposite."} 195 | {"id": "98", "hypothesis": "It appeared to him that she had betrayed him.", "label": "Entailment", "premise": "It occured to him that she has betrayed him.", "explanation": "The phrase it appeared to him means that he had a sudden realization, so, the realization is that she has been betraying him."} 196 | {"id": "98", "hypothesis": "It appeared to him that she had betrayed him.", "label": "Contradiction", "premise": "It did not appear to him him that she is not always loyal to him.", "explanation": "Here the man appears to have been betrayed by the woman, while he still believes that she is loyal to him."} 197 | {"id": "99", "hypothesis": "She fell asleep to the sound of music drifting upwards from the drawing-room .", "label": "Entailment", "premise": "The music from the drawing room drifted upwards and she felt like asleep.", "explanation": "The phrase drifting upwards in this context means that the sound of the music was coming from the drawing room and slowly getting louder."} 198 | {"id": "99", "hypothesis": "She fell asleep to the sound of music drifting upwards from the drawing-room .", "label": "Contradiction", "premise": "She fell asleep to the sound of music sinking out from the drawing-room .", "explanation": "In the one statement, the sound of music is drifting upwards whereas in in the contradiction, the music is sinking out."} 199 | {"id": "100", "hypothesis": "His playing secured him as a Romantic.", "label": "Entailment", "premise": "His playing is an epitome of Romantic.", "explanation": "Romanticism was a movement in the arts and literature that began in the late 18th century and It emphasized imagination, emotion, and individualism, here, the word secured is used to mean that his playing style resemble the romantic movement."} 200 | {"id": "100", "hypothesis": "His playing secured him as a Romantic.", "label": "Contradiction", "premise": "His playing denounced him as a Romantic.", "explanation": "To secure something means to make it certain or confirm it, while to denounce something means to speak out against it or condemn it."} 201 | {"id": "101", "hypothesis": "The promise between us was a flower.", "label": "Entailment", "premise": "The promise between us was beautiful and fragile.", "explanation": "A flower is something that is beautiful and fragile, sometime a promise could be similar."} 202 | {"id": "101", "hypothesis": "The promise between us was a flower.", "label": "Contradiction", "premise": "The promise between us was a poisonous snake.", "explanation": "A flower is something that is beautiful and represents growth, while a poisonous snake is something that is dangerous and can harm."} 203 | {"id": "102", "hypothesis": "A golden sun shines high in the sky.", "label": "Entailment", "premise": "A very bright sun shines high in the sky.", "explanation": "The sun is often described as being golden in color, and it is very bright when it is high in the sky."} 204 | {"id": "102", "hypothesis": "A golden sun shines high in the sky.", "label": "Contradiction", "premise": "A sunset is setting in the west.", "explanation": "A sunset is the time when the sun sets in the west, so it is not possible for a golden sun to be shining high in the sky at the same time."} 205 | {"id": "103", "hypothesis": "The stars cha-cha in the night sky.", "label": "Entailment", "premise": "All the starts were dancing in the night sky.", "explanation": "Cha-cha means to dance, thus, this metaphor meaning that the stars are dancing in the night sky."} 206 | {"id": "103", "hypothesis": "The stars cha-cha in the night sky.", "label": "Contradiction", "premise": "Stars start fading in the night sky.", "explanation": "The stars cha-cha in the night sky would suggest that they are shining brightly and are very visible, while stars fading when the sky turns dark would suggest that they are becoming dimmer and less visible."} 207 | {"id": "104", "hypothesis": "Blazed and flared the red horizon ,", "label": "Entailment", "premise": "Blazed and shone out suddenly the red horizon ,", "explanation": "To flare is to burn brightly and suddenly."} 208 | {"id": "104", "hypothesis": "Blazed and flared the red horizon ,", "label": "Contradiction", "premise": "Blazed and dimmed out the red horizon ,", "explanation": "To flare is to burn brightly and suddenly where dimmed out means the exact opposite."} 209 | {"id": "105", "hypothesis": "He felt a wave of anger.", "label": "Entailment", "premise": "He feels a seething anger inside.", "explanation": "A seething anger is a feeling of anger that is intense like a wave."} 210 | {"id": "105", "hypothesis": "He felt a wave of anger.", "label": "Contradiction", "premise": "He felt mild anger.", "explanation": "It is not possible to feel intense anger or a wave of anger and mild anger at the same time."} 211 | {"id": "106", "hypothesis": "I was alone in a sea of unknown faces.", "label": "Entailment", "premise": "I felt really alone when I was surrounded by all these people I didn't know.", "explanation": "Being surrounded by unknown faces mean people you don't know are surrounding you, which can make you feel alone because you have no one to talk to."} 212 | {"id": "106", "hypothesis": "I was alone in a sea of unknown faces.", "label": "Contradiction", "premise": "I was alone amidst many familiar faces.", "explanation": "Being alone amidst familiar faces contradicts being alone in sea of unknown faces who are primarily not familiar."} 213 | {"id": "107", "hypothesis": "I miss wandering in her eyes and watching her.", "label": "Entailment", "premise": "I miss staring in her eyes and watching her.", "explanation": "Wandering in her eyes means looking into her eyes and watching her."} 214 | {"id": "107", "hypothesis": "I miss wandering in her eyes and watching her.", "label": "Contradiction", "premise": "I miss neglecting her eyes and watching her.", "explanation": "To wander in something means to explore it or move about freely, while to neglect something means to disregard it or pay no attention to it."} 215 | {"id": "108", "hypothesis": "All fire-flush 'd when forest trees crumbled.", "label": "Entailment", "premise": "All fire-flush 'd when forest trees smoldered.", "explanation": "Crumble means to collapse or break into pieces, typically because of structural damage or decay."} 216 | {"id": "108", "hypothesis": "All fire-flush 'd when forest trees crumbled.", "label": "Contradiction", "premise": "All fire-flush 'd when forest trees grew.", "explanation": "All fire-flush'd describes a situation where everything is on fire and burning, i.e., the trees are crumbled while all forest trees grew describes a situation where everything is growing and flourishing."} 217 | {"id": "109", "hypothesis": "He heaped him with work.", "label": "Entailment", "premise": "He burdened him with work.", "explanation": "To heap means to pile something high, so the person is giving a lot of work."} 218 | {"id": "109", "hypothesis": "He heaped him with work.", "label": "Contradiction", "premise": "He rewarded him with relaxation at his job.", "explanation": "To heap something on someone means to burden them with it, while to reward someone means to give them something for their efforts."} 219 | {"id": "110", "hypothesis": "I am the good shepherd and I lay down my life for the sheep.", "label": "Entailment", "premise": "I am such a good person that others are easily taking advantage of me.", "explanation": "In this context, the speaker is saying that they are the good shepherd who takes care of the others while others can take advantage of their selflessness."} 220 | {"id": "110", "hypothesis": "I am the good shepherd and I lay down my life for the sheep.", "label": "Contradiction", "premise": "I am the bad shepherd and I take the sheep to a place where they are slaughtered for my own pleasure.", "explanation": "In the other statement, the speaker is likely a person who enjoys causing harm to others, and sees themselves as the bad shepherd who takes their victims to a place where they will be hurt or killed, whereas the one statement which states the person is a good shephard."} 221 | {"id": "111", "hypothesis": "The glut of oil depressed gas prices.", "label": "Entailment", "premise": "The over-production of oil at market depressed gasoline prices.", "explanation": "The glut of oil means there was too much oil and it caused the gas prices to go down."} 222 | {"id": "111", "hypothesis": "The glut of oil depressed gas prices.", "label": "Contradiction", "premise": "The glut of oil raised gas prices.", "explanation": "The word glut signifies an excess or surplus of something, while depressed typically connotes a lowering or decrease in price, in this instance, the contradictory sentence is saying that the excess of oil actually caused prices to go up, which contradicts the one statement."} 223 | {"id": "112", "hypothesis": "And his heart outran his footsteps.", "label": "Entailment", "premise": "His heart runs faster than his footsteps.", "explanation": "outran means to go faster than something else, in this case, his heart is going faster than his footsteps."} 224 | {"id": "112", "hypothesis": "And his heart outran his footsteps.", "label": "Contradiction", "premise": "And his heart raced slower than his feet could carry him.", "explanation": "When someone's heart outran their footsteps, it means they were running so fast that their heart was racing ahead of them, which is contradictory in the other statement."} 225 | {"id": "113", "hypothesis": "The government floated the ruble for a few months.", "label": "Entailment", "premise": "The government allowed the value of the ruble to be determined by the markets for a few months.", "explanation": "Floating the ruble means that the government allowed the value of the currency to be determined by supply and demand instead of pegging it to another currency or a set value."} 226 | {"id": "113", "hypothesis": "The government floated the ruble for a few months.", "label": "Contradiction", "premise": "the government immediately sunk the ruble.", "explanation": "To float something means to support or keep it afloat, while to sink something means to let it sink or allow it to fail."} 227 | {"id": "115", "hypothesis": "My son drifted around for years in california before going to law school.", "label": "Entailment", "premise": "My son drifted around all over the place, wasting his time in California, before settling down in law school.", "explanation": "My son drifted around means he didn't have a specific goal or plan and just wasted his time before finally settling down and going to law school."} 228 | {"id": "115", "hypothesis": "My son drifted around for years in california before going to law school.", "label": "Contradiction", "premise": "All my son has ever done is live at home and go to law school.", "explanation": "Drifted typically connotes aimlessness or going from place to place while is live at home connotes that the son stayed at home all the time before joining the law school."} 229 | {"id": "116", "hypothesis": "The night sky blazed with the massive bombardment.", "label": "Entailment", "premise": "There was a big hell broke out in night sky.", "explanation": "The night sky blazed with the massive bombardment means that there was a big explosion in the sky."} 230 | {"id": "116", "hypothesis": "The night sky blazed with the massive bombardment.", "label": "Contradiction", "premise": "The night sky became lighter and brighter because of the excessive bombardment .", "explanation": "To blaze with something means to be filled or covered with it, while to become lighter and brighter because of something would suggest the opposite."} 231 | {"id": "117", "hypothesis": "Then a snow-white cloud unfolding ,", "label": "Entailment", "premise": "Then a snow-white cloud developing ,", "explanation": "Unfolding means to develop or grow here."} 232 | {"id": "117", "hypothesis": "Then a snow-white cloud unfolding ,", "label": "Contradiction", "premise": "Then a snow-white cloud shrinked.", "explanation": "Unfolding means to develop or grow here whereas shrinking is used to mean the cloud is reducing."} 233 | {"id": "118", "hypothesis": "She sprinkled sunflower seeds.", "label": "Entailment", "premise": "Sunflower seeds were sown by her.", "explanation": "Sprinkling sunflower seeds means that she scattered them on the ground."} 234 | {"id": "118", "hypothesis": "She sprinkled sunflower seeds.", "label": "Contradiction", "premise": "She assembled sunflower seeds.", "explanation": "Sprinkling sunflower seeds would suggest that she scattered them on the ground while assembling means she is collecting the seeds."} 235 | {"id": "119", "hypothesis": "A flame ablaze that skipped amid the skies.", "label": "Entailment", "premise": "A flame ablaze that appear amid the skies.", "explanation": "A flame ablaze that appear amid the skies means a flame that appears in the sky."} 236 | {"id": "119", "hypothesis": "A flame ablaze that skipped amid the skies.", "label": "Contradiction", "premise": "I see flames ablaze and then disappear after a while in the sky.", "explanation": "A flame ablaze that skipped amid the skies would suggest that the fire was burning brightly and leaping up into the air, while flames ablaze and then disappear after a while in the sky would suggest that the fire was burning and then extinguished."} 237 | {"id": "120", "hypothesis": "Dashed by the refusal.", "label": "Entailment", "premise": "Disheartened by the refusal.", "explanation": "The person is disappointed because their request was refused."} 238 | {"id": "120", "hypothesis": "Dashed by the refusal.", "label": "Contradiction", "premise": "Remained cheerful at the rejection.", "explanation": "In the one statement the person is disappointed while the other statement says the person is still cheerful."} 239 | {"id": "121", "hypothesis": "As the springs to meet the sunshine ,", "label": "Entailment", "premise": "As the springs to be exposed to the sunshine ,", "explanation": "Meeting the sun is seen as being exposed to the sunshine."} 240 | {"id": "121", "hypothesis": "As the springs to meet the sunshine ,", "label": "Contradiction", "premise": "As the springs never meet the sunshine ,", "explanation": "As the springs to meet the sunshine would suggest that they are being exposed to the sunshine which is contradicted by the other statement ."} 241 | {"id": "122", "hypothesis": "Life in the camp wilted him.", "label": "Entailment", "premise": "Living in a camp for weeks tired him.", "explanation": "Wilted means to become weak or feeble, so, the person is becoming weak from living in the camp for too long."} 242 | {"id": "122", "hypothesis": "Life in the camp wilted him.", "label": "Contradiction", "premise": "The camp life really had energized him.", "explanation": "To wilt means to lose vitality or strength, while to energize means to give strength or energy to."} 243 | {"id": "123", "hypothesis": "That boosted Raavan's spirits.", "label": "Entailment", "premise": "The demon king Raavan became happy.", "explanation": "When someone becomes happy, their spirits are boosted."} 244 | {"id": "123", "hypothesis": "That boosted Raavan's spirits.", "label": "Contradiction", "premise": "That brought Raavan down.", "explanation": "Raavan's spirits may have been boosted by some fact and at the same time the spirit cannot brought down."} 245 | {"id": "124", "hypothesis": "He peppered fertilizer over the lawn.", "label": "Entailment", "premise": "He was doing some gardening.", "explanation": "The person is using a metaphor to describe how he is spreading the chemicals typically used for gardening."} 246 | {"id": "124", "hypothesis": "He peppered fertilizer over the lawn.", "label": "Contradiction", "premise": "He gathered fertilizer over the lawn.", "explanation": "To pepper something means to sprinkle it lightly, while to gather something means to collect it together."} 247 | {"id": "125", "hypothesis": "She was just a trophy to Ricardo.", "label": "Entailment", "premise": "Ricardo saw her as nothing more than an accomplishment.", "explanation": "A trophy is something that is won or achieved, usually as a sign of victory."} 248 | {"id": "125", "hypothesis": "She was just a trophy to Ricardo.", "label": "Contradiction", "premise": "She was the love of Ricardo's life.", "explanation": "A trophy is something that is won or achieved, while the love of one's life is someone who is cherished and loved deeply."} 249 | -------------------------------------------------------------------------------- /testgolddata/simile_test.jsonl: -------------------------------------------------------------------------------- 1 | {"id": 2, "premise": "The place looked impenetrable and inescapable", "hypothesis": "The place looked like a fortress", "label": "Entailment", "explanation": "A fortress is a military stronghold, hence it would be very hard to walk into, or in other words impenetrable."} 2 | {"id": 2, "premise": "The place looked easy to walk into and exit from", "hypothesis": "The place looked like a fortress", "label": "Contradiction", "explanation": "A fortress is a military stronghold, hence it would be very hard to walk into and exit from it."} 3 | {"id": 7, "premise": "Rays of sun diaphanously pierced through the leaves", "hypothesis": "Scattered sunlight poked through the leaves like a moth-eaten curtain .", "label": "Entailment", "explanation": "A moth-eaten curtain is a curtain with holes in it, so the sun is poking through leaves in a similar way, in that it is coming through in patches in a light, delicate, and translucent manner."} 4 | {"id": 7, "premise": "Rays of sun thickly and densely covered the leaves.", "hypothesis": "Scattered sunlight poked through the leaves like a moth-eaten curtain .", "label": "Contradiction", "explanation": "A moth-eaten curtain would have holes in it, allowing light to come through in a scattered and delicate way, rather than in a thick and dense way"} 5 | {"id": 10, "premise": "It was rapidly becoming smaller and smaller on the horizon, turning into nothing but a sliver.", "hypothesis": "It was becoming nothing but a sliver on the horizon looking like a closing eye .", "label": "Entailment", "explanation": "An eye that is closing is rapidly becoming smaller, and thus the simile is implying that the object on the horizon is rapidly shrinking."} 6 | {"id": 10, "premise": "It was rapidly becoming larger and larger on the horizon, turning into sliver.", "hypothesis": "It was becoming nothing but a sliver on the horizon looking like a closing eye .", "label": "Contradiction", "explanation": "An eye that is closing would be getting smaller, not larger"} 7 | {"id": 23, "premise": "He appeared carefree", "hypothesis": "He appeared like a father tucking in his kids.", "label": "Entailment", "explanation": "A father tucking in his kids is a carefree and content image."} 8 | {"id": 23, "premise": "He appeared burdened.", "hypothesis": "He appeared like a father tucking in his kids.", "label": "Contradiction", "explanation": "A father tucking in his kids is a loving image, whereas a burdened person is someone who is overwhelmed with responsibilities"} 9 | {"id": 30, "premise": "Whatever part of him that was remotely attractive has now become putrid.", "hypothesis": "Really, whatever part of him was decently cute or hot has suddenly become like a rotten sulfuric swamp .", "label": "Entailment", "explanation": "A rotten sulfuric swamp is a place that is incredibly rank and foul-smelling, so the comparison is emphasizing just how unattractive the person has become."} 10 | {"id": 30, "premise": "Whatever part of him that was remotely attractive has now become more beautiful.", "hypothesis": "Really, whatever part of him was decently cute or hot has suddenly become like a rotten sulfuric swamp .", "label": "Contradiction", "explanation": "A rotten sulfuric swamp is a place that is incredibly smelly and full of toxic gas, so saying something is like a rotten sulfuric swamp implies it is incredibly unattractive"} 11 | {"id": 33, "premise": "Even though the fall colors were rich and striking, her spirits were still low.", "hypothesis": "The fall colors were like a landscape painting, but even that didnt lift her spirits.", "label": "Entailment", "explanation": "A landscape painting is usually seen as a beautiful thing with rich vibrant colors, but even that couldn't cheer her up, indicating that her spirits were very low."} 12 | {"id": 33, "premise": "Even though the fall colors were dull and uninteresting, her spirits were still low.", "hypothesis": "The fall colors were like a landscape painting, but even that didnt lift her spirits.", "label": "Contradiction", "explanation": "A landscape painting is typically colorful and full of life, whereas dull and uninteresting colors would not be found in a painting"} 13 | {"id": 39, "premise": "It was astoundingly refreshing", "hypothesis": "It was like an english rain shower just when one was overly warm from a long ride through the countryside.", "label": "Entailment", "explanation": "An english rain shower is sudden and refreshing, which is what the speaker found the event to be."} 14 | {"id": 39, "premise": "It was incredibly weary.", "hypothesis": "It was like an english rain shower just when one was overly warm from a long ride through the countryside.", "label": "Contradiction", "explanation": "A rain shower is refreshing, so it would not make someone incredibly weary"} 15 | {"id": 49, "premise": "Just the thought of it made Dana feel flushed and embarrassed.", "hypothesis": "Thinking about it made dana feel like a giggly girl", "label": "Entailment", "explanation": "A giggly girl is someone who is easily amused or excited, and is also generally flushed embarrassed easily."} 16 | {"id": 49, "premise": "Just the thought of it made Dana feel cool and collected.", "hypothesis": "Thinking about it made dana feel like a giggly girl", "label": "Contradiction", "explanation": "Giggly girls are usually not cool and collected, they are more likely to be silly and carefree"} 17 | {"id": 54, "premise": "The smell was noxious .", "hypothesis": "The smell was like a clogged sewer .", "label": "Entailment", "explanation": "A clogged sewer is a place where waste accumulates and emits a strong, unpleasant smell."} 18 | {"id": 54, "premise": "The smell was pleasant.", "hypothesis": "The smell was like a clogged sewer .", "label": "Contradiction", "explanation": "A clogged sewer is a very unpleasant smell, so it would not be considered pleasant"} 19 | {"id": 63, "premise": "His features were rough, he had a few scars on his face, and his hair was cut very short.", "hypothesis": "He had rough features, his face bearing a few scars, and his hair was cut like a military haircut", "label": "Entailment", "explanation": "A military haircut is very short and conservative, so this simile is emphasizing that his hair was cut very short."} 20 | {"id": 63, "premise": "His features were rough, he had a few scars on his face, and his hair was cut very long.", "hypothesis": "He had rough features, his face bearing a few scars, and his hair was cut like a military haircut", "label": "Contradiction", "explanation": "A military haircut is very short, so saying it was cut like a military haircut means it was also cut short"} 21 | {"id": 65, "premise": "The swamp was silent", "hypothesis": "The swamp was like a huge, never-ending cemetery", "label": "Entailment", "explanation": "A cemetery is a place where people go to mourn the dead, and it is typically very silent."} 22 | {"id": 65, "premise": "The swamp was loud.", "hypothesis": "The swamp was like a huge, never-ending cemetery", "label": "Contradiction", "explanation": "A cemetery is a place that is meant to be peaceful and quiet, not loud"} 23 | {"id": 66, "premise": "She was flexible", "hypothesis": "She was like a modelling clay for children.", "label": "Entailment", "explanation": "Children can mold clay into any shape they want, so the simile is implying that the woman is flexible."} 24 | {"id": 66, "premise": "She was inflexible and rigid.", "hypothesis": "She was like a modelling clay for children.", "label": "Contradiction", "explanation": "Modelling clay is soft, pliable, and easy to mould, whereas inflexible and rigid things are not"} 25 | {"id": 72, "premise": "Even though most of my previous boyfriends were bad boys, I held onto the hope that beneath their exterior they were actually sweet and gushy.", "hypothesis": "Even though I hadnt had a lot of good fortune with boyfriends, I held on to the hope that beneath their bad boy, brash exterior they were like a chocolate lava cake", "label": "Entailment", "explanation": "A chocolate lava cake is a dessert that is gooey and sweet on the inside, even though it may have a hard exterior."} 26 | {"id": 72, "premise": "Even though most of my previous boyfriends were bad boys, I held onto the hope that beneath their exterior they were actually bitter and angry.", "hypothesis": "Even though I hadnt had a lot of good fortune with boyfriends, I held on to the hope that beneath their bad boy, brash exterior they were like a chocolate lava cake", "label": "Contradiction", "explanation": "A chocolate lava cake is a dessert that is soft and gooey on the inside, whereas a bitter and angry person would be the opposite"} 27 | {"id": 76, "premise": "He asks, his voice soft and mellifluous", "hypothesis": "He asks, his voice like a bbc broadcaster", "label": "Entailment", "explanation": "A BBC broadcaster has a professional, calming voice which is why the simile is used."} 28 | {"id": 76, "premise": "He asks, his voice harsh and grating.", "hypothesis": "He asks, his voice like a bbc broadcaster", "label": "Contradiction", "explanation": "A BBC broadcaster typically has a very cultivated voice, so saying his voice was like that implies it was not harsh and grating"} 29 | {"id": 77, "premise": "We had to hurry across the lawn to make it to the exhibition on time, and by then, Nolan's symphony was in its fourth movement which was loud, percussive and quick.", "hypothesis": "By the time wed finally crossed the lawn to the exhibition, nolans symphony was in its fourth movement: like a sadistic march .", "label": "Entailment", "explanation": "A sadistic march is a march that is cruel and punishing, which implies that it was loud, percussive, and quick."} 30 | {"id": 77, "premise": "We had to hurry across the lawn to make it to the exhibition on time, and by then, Nolan's symphony was in its fourth movement which was soft, mellow and slow.", "hypothesis": "By the time wed finally crossed the lawn to the exhibition, nolans symphony was in its fourth movement: like a sadistic march .", "label": "Contradiction", "explanation": "A sadistic march would be a fast tempo, and not soft, mellow, and slow"} 31 | {"id": 78, "premise": "That day and night passed without event and time was seamless.", "hypothesis": "That day and night passed without event and time was like the clouds above.", "label": "Entailment", "explanation": "Similar to the way clouds are constantly moving and changing, time seemed to be moving and changing without any events happening."} 32 | {"id": 78, "premise": "That day and night passed without event and time was choppy and inconsistent.", "hypothesis": "That day and night passed without event and time was like the clouds above.", "label": "Contradiction", "explanation": "The clouds are generally seen as slow-moving and consistent, so time passing like the clouds would imply it was slow-moving and consistent, not choppy and inconsistent"} 33 | {"id": 79, "premise": "The moment was protracted and stiff as we stared at each other.", "hypothesis": "We stared at each other; the moment was like a heavily starched shirt .", "label": "Entailment", "explanation": "A heavily starched shirt is one that is protracted and stiff and comparison is implying that the moment was not very comfortable."} 34 | {"id": 79, "premise": "The moment was brief and relaxed as we stared at each other.", "hypothesis": "We stared at each other; the moment was like a heavily starched shirt .", "label": "Contradiction", "explanation": "A heavily starched shirt is uncomfortable and stiff, so the moment being like that would imply that it was uncomfortable and not relaxed"} 35 | {"id": 81, "premise": "They were sharing experiences that were molding them into the women they would become and because of this they were close.", "hypothesis": "They were sharing experiences that were molding them into the women they would become and because of this they were like a latticed climbing vine .", "label": "Entailment", "explanation": "A latticed climbing vine is a plant that grows close to other plants in order to receive support."} 36 | {"id": 81, "premise": "They were sharing experiences that were molding them into the women they would become and because of this they were distant.", "hypothesis": "They were sharing experiences that were molding them into the women they would become and because of this they were like a latticed climbing vine .", "label": "Contradiction", "explanation": "A latticed climbing vine implies that they were close to each other, not distant"} 37 | {"id": 84, "premise": "Our time is short,\" said Abner.", "hypothesis": "Said abner, our time is like a daggers edge .", "label": "Entailment", "explanation": "A dagger's edge is very thin and sharp, and typically very short and fleeting"} 38 | {"id": 84, "premise": "Our time is long,\" said Abner.", "hypothesis": "Said abner, our time is like a daggers edge .", "label": "Contradiction", "explanation": "A dagger's edge is very thin and sharp, meaning that it can cut easily, so saying that time is like a dagger's edge would imply that it is fleeting and can end quickly"} 39 | {"id": 86, "premise": "The swelling had gone down and the bandage around it felt strong and comfortable.", "hypothesis": "The swelling had gone down, and the bandage around it like a firm hand .", "label": "Entailment", "explanation": "A firm hand would be one that is strong and comfortable."} 40 | {"id": 86, "premise": "The swelling had gone down and the bandage around it felt weak and uncomfortable.", "hypothesis": "The swelling had gone down, and the bandage around it like a firm hand .", "label": "Contradiction", "explanation": "A firm hand would imply that the bandage is tight and secure, whereas a weak and uncomfortable bandage would imply the opposite"} 41 | {"id": 97, "premise": "The animal looked like a small German shepherd, but with a reddish-brown coat and a black mask over its eyes.", "hypothesis": "With its black mask and those ears, the animal was like a small german shepherd, but the rest of its fur was like a chestnut mare's .", "label": "Entailment", "explanation": "A chestnut mare is a horse with a reddish brown coat, so saying the animal's fur was like a chestnut mare's would mean it would have a reddish brown coat"} 42 | {"id": 97, "premise": "The animal looked like a small German shepherd, but with a blondish-white coat and a black mask over its eyes.", "hypothesis": "With its black mask and those ears, the animal was like a small german shepherd, but the rest of its fur was like a chestnut mare's .", "label": "Contradiction", "explanation": "A chestnut mare is a horse with a reddish brown coat, so saying the animal's fur was like a chestnut mare's would mean it would have a reddish brown coat, not a blondish-white coat"} 43 | {"id": 101, "premise": "It felt soft and slimy .", "hypothesis": "It felt like a portobello mushroom .", "label": "Entailment", "explanation": "A portobello mushroom is a type of mushroom that is soft and slimy."} 44 | {"id": 101, "premise": "It felt hard and dry.", "hypothesis": "It felt like a portobello mushroom .", "label": "Contradiction", "explanation": "Portobello mushrooms are big and wet, so saying it felt like one would imply it felt slimy, not hard and dry"} 45 | {"id": 109, "premise": "Her smile was quiet and gentle .", "hypothesis": "Her smile was like a slow rising tide .", "label": "Entailment", "explanation": "A slow rising tide is a gentle and gradual increase, so her smile was likely gentle and gradual as well."} 46 | {"id": 109, "premise": "Her smile was loud and harsh.", "hypothesis": "Her smile was like a slow rising tide .", "label": "Contradiction", "explanation": "A slow rising tide is gentle and gradual, whereas a loud and harsh smile would be the opposite"} 47 | {"id": 118, "premise": "The beauty of it was breathtaking", "hypothesis": "The beauty of it was like a lost paradise", "label": "Entailment", "explanation": "A lost paradise is a place that is so beautiful and perfect that it seems unreal, hence it would take your breath away."} 48 | {"id": 118, "premise": "The beauty of it was uninspiring", "hypothesis": "The beauty of it was like a lost paradise", "label": "Contradiction", "explanation": "A lost paradise is a place that is incredibly beautiful, so saying that the beauty was uninspiring contradicts the simile"} 49 | {"id": 120, "premise": "His eyes were an alarming gray-green", "hypothesis": "His eyes were like a marvelous rain forest that had been blanketed with smoke", "label": "Entailment", "explanation": "A rain forest is lush and green, whereas smoke is gray and when a rainforest is covered with a blanket of smoke its an alarming gray green color"} 50 | {"id": 120, "premise": "His eyes were a calming blue.", "hypothesis": "His eyes were like a marvelous rain forest that had been blanketed with smoke", "label": "Contradiction", "explanation": "A rainforest is usually a very vibrant and green place, but if it is blanketed with smoke, it would be a alarmingly gray green color"} 51 | {"id": 133, "premise": "It was gentle", "hypothesis": "It was like a final tendril of smoke from a dying candle.", "label": "Entailment", "explanation": "A tendril of smoke is thin and light, so it would be gentle."} 52 | {"id": 133, "premise": "It was harsh", "hypothesis": "It was like a final tendril of smoke from a dying candle.", "label": "Contradiction", "explanation": "The smoke from a dying candle is usually thin since the candle is about to end, and not harsh"} 53 | {"id": 155, "premise": "He took my hand in his, and his movements were gentle.", "hypothesis": "He took hold of my hand, his every motion was like a sweet caress .", "label": "Entailment", "explanation": "A caress is a gentle touch, so the simile is saying that his every motion was gentle."} 54 | {"id": 155, "premise": "He took my hand in his, and his movements were forceful.", "hypothesis": "He took hold of my hand, his every motion was like a sweet caress .", "label": "Contradiction", "explanation": "A caress is a gentle touch or stroke, whereas forceful movements would imply he was not being gentle"} 55 | {"id": 158, "premise": "Three years of lessons come into my head all at once, bumping into each other.", "hypothesis": "Three years of lessons crowd my head like an unsettled flock of pigeons.", "label": "Entailment", "explanation": "A flock of pigeons is a large group of birds that are often seen bumping into each other."} 56 | {"id": 158, "premise": "Three years of lessons come into my head one at a time, in a linear fashion.", "hypothesis": "Three years of lessons crowd my head like an unsettled flock of pigeons.", "label": "Contradiction", "explanation": "Pigeons tend to fly in flocks, and are not linear in their movements, so an unsettled flock of pigeons implies that the lessons are not in a linear fashion"} 57 | {"id": 164, "premise": "Her eyebrows were too thick and black, and her teeth looked crooked and yellow.", "hypothesis": "Her eyebrows were drawn in too thick and black, and her teeth like a sepia image of a rickety fence.", "label": "Entailment", "explanation": "A sepia image is usually yellow in tone and a rickety fence is crooked so a teeth like sepia image of a rickety fence is yellow and crooked."} 58 | {"id": 164, "premise": "Her eyebrows were too thick and black, and her teeth looked straight and white.", "hypothesis": "Her eyebrows were drawn in too thick and black, and her teeth like a sepia image of a rickety fence.", "label": "Contradiction", "explanation": "A sepia image is an old, faded photograph yellow in color and a rickety fence is crooked, so her teeth looking like a sepia image of a rickety fence would mean they are crooked and yellow, not straight and white"} 59 | {"id": 174, "premise": "Annie seemed limp .", "hypothesis": "Annie seemed like a cast- aside rag doll .", "label": "Entailment", "explanation": "A rag doll is an inanimate object that is not able to move on its own, so it would appear limp."} 60 | {"id": 174, "premise": "Annie seemed full of energy.", "hypothesis": "Annie seemed like a cast- aside rag doll .", "label": "Contradiction", "explanation": "A rag doll is usually dirty and tattered, and not full of energy"} 61 | {"id": 181, "premise": "The tunnels were black, and the only light came from the dim lantern we carried.", "hypothesis": "Other than the dim light we carried only because it was stuck to us, the tunnels were like a coal mine during a power outage.", "label": "Entailment", "explanation": "A coal mine during a power outage would be very dark, since there would be no lights on."} 62 | {"id": 181, "premise": "The tunnels were white, and the only light came from the dim lantern we carried.", "hypothesis": "Other than the dim light we carried only because it was stuck to us, the tunnels were like a coal mine during a power outage.", "label": "Contradiction", "explanation": "Coal mines are pitch black, so if the tunnels were like a coal mine, they would be completely dark, not just dimly lit"} 63 | {"id": 204, "premise": "He looked shriveled as he sat in a hardback chair.", "hypothesis": "He was slouched in a hardback chair, looking like a deflated balloon.", "label": "Entailment", "explanation": "A deflated balloon is small and collapsed, so it implies that the person looks small and collapsed or shriveled in the chair."} 64 | {"id": 204, "premise": "He looked healthy and robust as he sat in a hardback chair.", "hypothesis": "He was slouched in a hardback chair, looking like a deflated balloon.", "label": "Contradiction", "explanation": "A deflated balloon is saggy and not full of air usually small and collapsed, whereas a healthy person would be the opposite"} 65 | {"id": 210, "premise": "I silently thank Tess as the blue pill warms me up from the inside out, soothing my nerves.", "hypothesis": "I say a silent thanks to tess-already the blue pill is warming me up from the inside out, like a bowl of hot soup on an icy evening.", "label": "Entailment", "explanation": "A bowl of hot soup on an icy evening is very comforting, so the blue pill must be very comforting as well."} 66 | {"id": 210, "premise": "I silently thank Tess as the blue pill warms me from the inside out, wrecking my nerves.", "hypothesis": "I say a silent thanks to tess-already the blue pill is warming me up from the inside out, like a bowl of hot soup on an icy evening.", "label": "Contradiction", "explanation": "A bowl of hot soup would be comforting on a cold evening since it would help a person to get warm, not make someone nervous"} 67 | {"id": 211, "premise": "The gesture was elegant", "hypothesis": "The gesture was like a ballet move", "label": "Entailment", "explanation": "A ballet move is an elegant and fluid movement."} 68 | {"id": 211, "premise": "The gesture was awkward.", "hypothesis": "The gesture was like a ballet move", "label": "Contradiction", "explanation": "A ballet move is supposed to be fluid and elegant, whereas an awkward gesture is usually jerky and not well-executed"} 69 | {"id": 213, "premise": "Lowering me to the floor, he hugged me close, making sure that he didn't release me until my shaking had stopped and I could stand on my own, although I felt wobbly", "hypothesis": "Lowering me to the floor, he hugged me close, making sure that he didn't release me until my shaking had stopped and I could stand on my own, although I felt like an infant.", "label": "Entailment", "explanation": "An infant is a young child who is not yet able to walk unaided, so the person must have felt wobbly when they were first released."} 70 | {"id": 213, "premise": "Lowering me to the floor, he hugged me close, making sure that he didn't release me until my shaking had stopped and I could stand on my own, although I felt sturdy.", "hypothesis": "Lowering me to the floor, he hugged me close, making sure that he didn't release me until my shaking had stopped and I could stand on my own, although I felt like an infant.", "label": "Contradiction", "explanation": "An infant is not sturdy, meaning they are not yet able to hold themselves up"} 71 | {"id": 218, "premise": "Even though she screamed and tried to hit him, she was defenseless.", "hypothesis": "She screamed and twisted and tried to hit him but she was like a chicken held upside down by its legs.", "label": "Entailment", "explanation": "A chicken held upside down by its legs is unable to move or defend itself in any way, making it helpless."} 72 | {"id": 218, "premise": "Even though she screamed and tried to hit him, she was well-protected.", "hypothesis": "She screamed and twisted and tried to hit him but she was like a chicken held upside down by its legs.", "label": "Contradiction", "explanation": "When chickens are held upside down by their legs, they are unable to move and are vulnerable to being killed, so the simile is implying that the woman is helpless and unprotected"} 73 | {"id": 220, "premise": "It felt hard", "hypothesis": "It felt like an iron bar.", "label": "Entailment", "explanation": "An iron bar would be very hard to the touch."} 74 | {"id": 220, "premise": "It felt soft.", "hypothesis": "It felt like an iron bar.", "label": "Contradiction", "explanation": "Iron is a hard metal, so saying it felt like an iron bar implies that it was hard, not soft"} 75 | {"id": 227, "premise": "The figure seemed blurred", "hypothesis": "The figure seemed like a smudge on a photograph negative.", "label": "Entailment", "explanation": "A smudge on a photograph negative will make the image appear blurred when it is developed."} 76 | {"id": 227, "premise": "The figure seemed sharp and clear", "hypothesis": "The figure seemed like a smudge on a photograph negative.", "label": "Contradiction", "explanation": "A smudge on a photograph is typically a dark, blurry area, so saying the figure was like a smudge on a photograph negative would mean it was dark and blurry, not sharp and clear"} 77 | {"id": 239, "premise": "His countenance was calm", "hypothesis": "His countenance was like a man strolling in the park on a warm summer's day.", "label": "Entailment", "explanation": "A person strolling in the park on a warm summer's day is usually calm and relaxed."} 78 | {"id": 239, "premise": "His countenance was agitated", "hypothesis": "His countenance was like a man strolling in the park on a warm summer's day.", "label": "Contradiction", "explanation": "A person who is strolling in the park on a warm summer's day is typically relaxed, not agitated"} 79 | {"id": 247, "premise": "The ray was especially known for its thick french toast.", "hypothesis": "The ray was famous for its french toast, which was like a brick", "label": "Entailment", "explanation": "A brick is a very thick object, so the french toast must be similarly thick."} 80 | {"id": 247, "premise": "The ray was especially known for its thin french toast.", "hypothesis": "The ray was famous for its french toast, which was like a brick", "label": "Contradiction", "explanation": "A brick is a thick, dense object, whereas thin toast is, well, thin"} 81 | {"id": 248, "premise": "And when I pulled up to the crooked iron gates that marked the boundaries of ravenwood, the air around us crackled and charged", "hypothesis": "And when I pulled up to the crooked iron gates that marked the boundaries of ravenwood, the air around us like an electrical storm.", "label": "Entailment", "explanation": "An electrical storm is characterized by a crackling and charged atmosphere."} 82 | {"id": 248, "premise": "And when I pulled up to the crooked iron gates that marked the boundaries of ravenwood, the air around us was still and calm.", "hypothesis": "And when I pulled up to the crooked iron gates that marked the boundaries of ravenwood, the air around us like an electrical storm.", "label": "Contradiction", "explanation": "An electrical storm is characterized by strong winds and heavy rains, whereas the air being still and calm would be the opposite"} 83 | {"id": 250, "premise": "But the hand that was pressing against my mouth was immovable.", "hypothesis": "But the hand pressing against my mouth was like a wall.", "label": "Entailment", "explanation": "A wall is an immovable object, so the simile is emphasizing that the hand was pressing very hard and was difficult to move."} 84 | {"id": 250, "premise": "But the hand that was pressing against my mouth was moving.", "hypothesis": "But the hand pressing against my mouth was like a wall.", "label": "Contradiction", "explanation": "A wall is a stationary object, whereas a hand is a movable object"} 85 | {"id": 254, "premise": "Her expression was solemn", "hypothesis": "Her expression was like a dirge.", "label": "Entailment", "explanation": "A dirge is a slow, somber song, usually sung at a funeral, so her expression was solemn."} 86 | {"id": 254, "premise": "Her expression was joyful.", "hypothesis": "Her expression was like a dirge.", "label": "Contradiction", "explanation": "A dirge is a slow, mournful song typically sung at a funeral, so it would not be joyful"} 87 | {"id": 259, "premise": "No, his menace was blatant", "hypothesis": "No, his menace was like a flashing neon sign.", "label": "Entailment", "explanation": "A flashing neon sign is a very obvious sign, so the simile is emphasizing how blatant the menace is."} 88 | {"id": 259, "premise": "No, his menace was subtle.", "hypothesis": "No, his menace was like a flashing neon sign.", "label": "Contradiction", "explanation": "A flashing neon sign is very bright and in-your-face, so it would be hard to miss"} 89 | {"id": 260, "premise": "Wielding strength and power was exhilarating for her.", "hypothesis": "The strength and power she wielded was like a drug.", "label": "Entailment", "explanation": "A drug is something that gives you a high, or in other words, is exhilarating."} 90 | {"id": 260, "premise": "Wielding strength and power was exhausting for her.", "hypothesis": "The strength and power she wielded was like a drug.", "label": "Contradiction", "explanation": "A drug is typically something that makes a person feel good and not exhausted"} 91 | {"id": 266, "premise": "Inside, I was giddy.", "hypothesis": "Inside, I was like a thirteen-year-old at a boy band pop concert.", "label": "Entailment", "explanation": "A thirteen-year-old at a boy band pop concert would be very excited and happy, or in other words, giddy."} 92 | {"id": 266, "premise": "Inside, I was solemn.", "hypothesis": "Inside, I was like a thirteen-year-old at a boy band pop concert.", "label": "Contradiction", "explanation": "A thirteen-year-old at a boy band pop concert would likely be screaming and excited to see a favorite artist, not solemn"} 93 | {"id": 281, "premise": "This place was quiet.", "hypothesis": "This place was like a funeral parlor.", "label": "Entailment", "explanation": "A funeral parlor is a place for grieving and silence, so it would be expected that it would be quiet."} 94 | {"id": 281, "premise": "This place was loud.", "hypothesis": "This place was like a funeral parlor.", "label": "Contradiction", "explanation": "A funeral parlor is a place for grieving, and is typically a somber place where people are quiet out of respect"} 95 | {"id": 286, "premise": "He felt like his heart was lacking something, or empty.", "hypothesis": "The truth was, he felt like his heart was like a politicians promise.", "label": "Entailment", "explanation": "A politicians promise is often not kept, or is broken, so the simile sentence is saying that his heart is lacking something."} 96 | {"id": 286, "premise": "He felt like his heart was full, or content.", "hypothesis": "The truth was, he felt like his heart was like a politicians promise.", "label": "Contradiction", "explanation": "A politician's promise is often broken, or not kept, in other words an empty promise, so saying someone's heart is like a politician's promise implies that the person's heart is empty"} 97 | {"id": 290, "premise": "By the time most of the seniors had left for the summer, the campus was deserted.", "hypothesis": "With a week left in the school year most of the seniors werent around campus anymore, their great railing was like a ski resort in summer.", "label": "Entailment", "explanation": "A ski resort in summer is a place that is typically devoid of people, so the simile is indicating that the campus was similarly empty."} 98 | {"id": 290, "premise": "By the time most of the seniors had left for the summer, the campus was bustling.", "hypothesis": "With a week left in the school year most of the seniors werent around campus anymore, their great railing was like a ski resort in summer.", "label": "Contradiction", "explanation": "A ski resort in summer would be pretty much abandoned because the snow melts during summer which prevents people from skiing"} 99 | {"id": 299, "premise": "Today he was sober", "hypothesis": "Today he was like a benedictine monk on the sabbath.", "label": "Entailment", "explanation": "A benedictine monk is someone who is sober and abstemious, so he is sober"} 100 | {"id": 299, "premise": "Today he was drunk.", "hypothesis": "Today he was like a benedictine monk on the sabbath.", "label": "Contradiction", "explanation": "A benedictine monk on the sabbath is someone who is supposed to be prayerful and meditative, not someone who is intoxicated"} 101 | {"id": 300, "premise": "His hair was gray and wispy", "hypothesis": "His hair was like a chemo patient", "label": "Entailment", "explanation": "Chemo patients often lose their hair, or have gray and wispy hair, so his hair was similar"} 102 | {"id": 300, "premise": "His hair was black and thick.", "hypothesis": "His hair was like a chemo patient", "label": "Contradiction", "explanation": "A chemo patient's hair is typically thin and may fall out due to the chemotherapy treatment, so black and thick hair would not be like a chemo patient's hair"} 103 | {"id": 302, "premise": "The mushroom cloud had dissipated, but the western sky was still angry pinkish-red.", "hypothesis": "The mushroom cloud was gone, but the western sky was like a bright weal of burnflesh.", "label": "Entailment", "explanation": "A weal is a raised, red mark on the skin, so the simile is suggesting that the sky looks like it has been angry displaying a pinkish red hue."} 104 | {"id": 302, "premise": "The mushroom cloud had dissipated, but the western sky was still calm and serene.", "hypothesis": "The mushroom cloud was gone, but the western sky was like a bright weal of burnflesh.", "label": "Contradiction", "explanation": "A weal is a raised, red mark on the skin, so the simile is suggesting that the sky looks like it has been angry displaying a pinkish red hue and not calm or serene"} 105 | {"id": 306, "premise": "On the third day, they woke up to a sky that looked like it was hanging low and was a dull gray color.", "hypothesis": "On the third day erin and dare woke to a sky that looked like an ominous shroud .", "label": "Entailment", "explanation": "A shroud is a cloth used to cover a dead body, and is often black or gray in color, so the sky was gray."} 106 | {"id": 306, "premise": "On the third day, they woke up to a sky that looked expansive and was a bright blue color.", "hypothesis": "On the third day erin and dare woke to a sky that looked like an ominous shroud .", "label": "Contradiction", "explanation": "A shroud is a piece of cloth used to cover a dead body, and is typically a dark color i.e black or gray and not blue"} 107 | {"id": 319, "premise": "The street system in London was contrived.", "hypothesis": "We were in london, where the street system was like a bowl of spaghetti.", "label": "Entailment", "explanation": "A bowl of spaghetti is a tangled mess, so the street system must be very convoluted and difficult to navigate."} 108 | {"id": 319, "premise": "The street system in London was simple and easy to navigate.", "hypothesis": "We were in london, where the street system was like a bowl of spaghetti.", "label": "Contradiction", "explanation": "A bowl of spaghetti is a mess of tangled noodles, implying that the streets are tangled and complicated, not simple"} 109 | {"id": 324, "premise": "She's powerless and weak", "hypothesis": "She's like a bee out of sting.", "label": "Entailment", "explanation": "A bee without a sting is powerless and weak, and cannot defend itself, so she is powerless."} 110 | {"id": 324, "premise": "She's powerful and strong", "hypothesis": "She's like a bee out of sting.", "label": "Contradiction", "explanation": "If a bee is out of sting, it means it is no longer a threat, and is weak, so saying she is like a bee out of sting would mean she is weak, not powerful and strong"} 111 | {"id": 328, "premise": "I looked over my shoulder with keen precision, but in the shadowy morning I was blind.", "hypothesis": "I looked over my shoulder with keen precision, but in the shadowy morning I was like a bat.", "label": "Entailment", "explanation": "Bats use echolocation to see in the dark, so the simile is saying that the speaker was looking around carefully but still couldn't see anything."} 112 | {"id": 328, "premise": "I looked over my shoulder with keen precision, and in the shadowy morning I could see everything.", "hypothesis": "I looked over my shoulder with keen precision, but in the shadowy morning I was like a bat.", "label": "Contradiction", "explanation": "Bats are nocturnal animals that use echolocation to navigate in the dark, so saying someone is like a bat in the shadowy morning would imply that they cannot see clearly"} 113 | {"id": 333, "premise": "The camp was silent", "hypothesis": "The camp was like a cemetery.", "label": "Entailment", "explanation": "A cemetery is a place where people go to mourn the dead, and it is usually very quiet, so the camp was silent."} 114 | {"id": 333, "premise": "The camp was loud", "hypothesis": "The camp was like a cemetery.", "label": "Contradiction", "explanation": "A cemetery is a place for the dead, and it is usually very quiet, so a camp being like a cemetery means it would be similarly quiet"} 115 | {"id": 336, "premise": "She felt empty, like someone had scooped her insides right out", "hypothesis": "She felt like a jack-o'-lantern.", "label": "Entailment", "explanation": "A jack-o'-lantern is a hollowed-out pumpkin, so the simile is emphasizing how empty she feels."} 116 | {"id": 336, "premise": "She felt full, like she had just eaten a hearty meal.", "hypothesis": "She felt like a jack-o'-lantern.", "label": "Contradiction", "explanation": "A jack-o'-lantern is a hollowed out pumpkin, so saying she felt like one would imply she felt empty, not full"} 117 | {"id": 342, "premise": "The giant milky membrane was sticky to touch.", "hypothesis": "I touched the giant milky membrane which was like a wet marshmallow .", "label": "Entailment", "explanation": "A wet marshmallow is a sticky food, so the simile is used to describe that the membrane feels sticky."} 118 | {"id": 342, "premise": "The giant milky membrane felt dry on touch.", "hypothesis": "I touched the giant milky membrane which was like a wet marshmallow .", "label": "Contradiction", "explanation": "A marshmallow is a type of candy that is soft and has a lot of moisture, so saying it is like a marshmallow would imply it is sticky"} 119 | {"id": 352, "premise": "She screamed loudly", "hypothesis": "She screamed like a fire truck siren", "label": "Entailment", "explanation": "A fire truck siren is very loud, so the simile is used to indicate that the woman was screaming very loudly."} 120 | {"id": 352, "premise": "She screamed softly.", "hypothesis": "She screamed like a fire truck siren", "label": "Contradiction", "explanation": "A siren is a loud, piercing sound, so a person screaming like a siren would be screaming loudly, not softly"} 121 | {"id": 355, "premise": "The forest was still dark", "hypothesis": "The forest was still like a high wall with serrated top.", "label": "Entailment", "explanation": "A wall with a serrated top would be jagged and foreboding, much like a dark forest."} 122 | {"id": 355, "premise": "The forest was still beaming with light", "hypothesis": "The forest was still like a high wall with serrated top.", "label": "Contradiction", "explanation": "A wall blocks out light, so it would be contradictory to say that the forest, which is like a wall, is beaming with light"} 123 | {"id": 357, "premise": "I had a clear vision from the start.", "hypothesis": "My vision from the beginning was like a well-pruned garden leaf .", "label": "Entailment", "explanation": "A well-pruned garden leaf is one that has been cared for and is healthy, in contrast to a leaf that has been neglected so the simile implies that speaker had a clear vision"} 124 | {"id": 357, "premise": "I had a blurry vision from the start", "hypothesis": "My vision from the beginning was like a well-pruned garden leaf .", "label": "Contradiction", "explanation": "A well-pruned garden leaf is trimmed and not blurry, so saying a vision is like a well-pruned garden leaf means it is clear and not blurry"} 125 | {"id": 362, "premise": "Today they were deep green .", "hypothesis": "Today they were like a freshly mowed football field .", "label": "Entailment", "explanation": "A freshly mowed football field is deep green."} 126 | {"id": 362, "premise": "Today they were pale and yellow.", "hypothesis": "Today they were like a freshly mowed football field .", "label": "Contradiction", "explanation": "A freshly mowed football field is green, not pale and yellow"} 127 | {"id": 368, "premise": "Juliana was studying him, and her blue eyes were clear.", "hypothesis": "Juliana was studying him, her eyes like a summer lake .", "label": "Entailment", "explanation": "A summer lake is usually calm, clear and serene, so the simile is saying that Juliana's eyes were clear as she was studying him."} 128 | {"id": 368, "premise": "Juliana was studying him, and her blue eyes were cloudy.", "hypothesis": "Juliana was studying him, her eyes like a summer lake .", "label": "Contradiction", "explanation": "A summer lake is typically clear and bright, whereas cloudy eyes are not"} 129 | {"id": 369, "premise": "We were munching next to the evening sun-swept grass and whispering trees of duboce park, where a bunch of happy dogs were sprinting around in circles while their owners were yapping at each other.", "hypothesis": "We stand there munching next to the evening sun-swept grass and whispering trees of duboce park, where a bunch of happy-ass dogs are sprinting around in circles while their owners stand there like a scarecrow convention .", "label": "Entailment", "explanation": "The image of a scarecrow convention implies that the people are not really paying attention to their surroundings, they are just yapping at each other."} 130 | {"id": 369, "premise": "We were munching next to the evening sun-swept grass and whispering trees of duboce park, where a bunch of happy dogs were sprinting around in circles while their owners were chatting amicably.", "hypothesis": "We stand there munching next to the evening sun-swept grass and whispering trees of duboce park, where a bunch of happy-ass dogs are sprinting around in circles while their owners stand there like a scarecrow convention .", "label": "Contradiction", "explanation": "A scarecrow convention would imply that the people are not happy, and are yapping each other, whereas the people in the park were chatting amicably"} 131 | {"id": 371, "premise": "Even though it was the middle of the day, the air was dark and had an odd grainy texture.", "hypothesis": "Though it was only mid-afternoon, the air was like an old newspaper photograph .", "label": "Entailment", "explanation": "An old newspaper photograph is dark and has a grainy texture, so the air must have been similarly dark and grainy."} 132 | {"id": 371, "premise": "Even though it was the middle of the day, the air was light and had a smooth texture.", "hypothesis": "Though it was only mid-afternoon, the air was like an old newspaper photograph .", "label": "Contradiction", "explanation": "An old photograph tends to be yellow and faded, whereas the air being light would imply it is not yellow"} 133 | {"id": 394, "premise": "The room was suddenly noisy and wild", "hypothesis": "The room was suddenly like a camp revival meeting", "label": "Entailment", "explanation": "A camp revival meeting is a religious gathering where people sing and dance energetically, so the simile sentence implies that the room became noisy and wild."} 134 | {"id": 394, "premise": "The room was suddenly quiet and calm", "hypothesis": "The room was suddenly like a camp revival meeting", "label": "Contradiction", "explanation": "A camp meeting would involve a lot of people so it is more likely to be chaotic and not quiet and calm"} 135 | {"id": 404, "premise": "Jodi looked paralyzed", "hypothesis": "Jodi looked like a stroke victim , trapped in her own mind, spittle forming in the corner of her mouth.", "label": "Entailment", "explanation": "A stroke victim is someone who has had a sudden paralysis, usually in the face or body, so it is implied that Jodi is paralyzed."} 136 | {"id": 404, "premise": "Jodi looked mobile and vibrant.", "hypothesis": "Jodi looked like a stroke victim , trapped in her own mind, spittle forming in the corner of her mouth.", "label": "Contradiction", "explanation": "A stroke victim is someone who has suffered damage to the brain, which can result in paralysis or loss of movement, so saying Jodi looked like a stroke victim contradicts saying she looked mobile and vibrant"} 137 | {"id": 405, "premise": "Sarah's convinced herself that I'm partly clairvoyant .", "hypothesis": "Sarah's convinced herself that I'm like a human magic 8 ball .", "label": "Entailment", "explanation": "A magic 8 ball is an object used for divination, which is a form of clairvoyance, so Sarah thinks the person is clairvoyant."} 138 | {"id": 405, "premise": "Sarah's convinced herself that I'm not at all clairvoyant.", "hypothesis": "Sarah's convinced herself that I'm like a human magic 8 ball .", "label": "Contradiction", "explanation": "Magic 8 balls are used as a fortune telling device, and are thus clairvoyant"} 139 | {"id": 412, "premise": "Despite that he loved rich food and ale, his stomach was still flat.", "hypothesis": "Despite his fondness for rich food and good brown ale, his stomach was like a board;", "label": "Entailment", "explanation": "A board is a flat surface, so despite his love for rich food, his stomach was still flat."} 140 | {"id": 412, "premise": "Despite that he loved rich food and ale, his stomach was still bloated.", "hypothesis": "Despite his fondness for rich food and good brown ale, his stomach was like a board;", "label": "Contradiction", "explanation": "A board is a flat, hard surface, so saying someone's stomach is like a board means it is flat and not bloated"} 141 | {"id": 422, "premise": "Tommy's eyes opened wide and he looked down at my stomach, which was flat.", "hypothesis": "Tommy's eyes opened really big and he glanced down at my stomach, which was like a pancake.", "label": "Entailment", "explanation": "A pancake is a flat, round object, so Tommy's eyes opening wide and looking down at my stomach implies that my stomach is flat."} 142 | {"id": 422, "premise": "Tommy's eyes opened wide and he looked down at my stomach, which was bloated.", "hypothesis": "Tommy's eyes opened really big and he glanced down at my stomach, which was like a pancake.", "label": "Contradiction", "explanation": "A pancake is a thin, flat cake, whereas a bloated stomach is distended and protruding"} 143 | {"id": 423, "premise": "Phillip wasn't brave enough to deal with something deadly.", "hypothesis": "Phillip wouldn't have had the guts to handle something like a scorpion.", "label": "Entailment", "explanation": "A scorpion is a deadly creature, so Phillip wouldn't have had the courage to deal with it."} 144 | {"id": 423, "premise": "Phillip wasn't brave enough to deal with something innocuous.", "hypothesis": "Phillip wouldn't have had the guts to handle something like a scorpion.", "label": "Contradiction", "explanation": "A scorpion is a dangerous and poisonous creature, it can even be deadly but it is not considered innocuous"} 145 | {"id": 426, "premise": "It was so quiet inside the house.", "hypothesis": "The inside of the house was like a tomb", "label": "Entailment", "explanation": "A tomb is a place where people are buried, and it is very quiet inside a tomb, implying the house was quiet."} 146 | {"id": 426, "premise": "It was so loud inside the house.", "hypothesis": "The inside of the house was like a tomb", "label": "Contradiction", "explanation": "A tomb is a place where people are buried, and it is typically very quiet inside"} 147 | {"id": 428, "premise": "The people around the fire pit had grown quiet.", "hypothesis": "The audience around the fire pit had grown like a graveyard.", "label": "Entailment", "explanation": "A graveyard is a place where people go to be silent and respectful, meaning the audience was silent."} 148 | {"id": 428, "premise": "The people around the fire pit had grown loud.", "hypothesis": "The audience around the fire pit had grown like a graveyard.", "label": "Contradiction", "explanation": "A graveyard is a place where people are silent out of respect for the dead, not loud"} 149 | {"id": 432, "premise": "She must have been fit", "hypothesis": "She must have been like a snow leopard.", "label": "Entailment", "explanation": "A snow leopard is a sleek and powerful animal, so the person must be fit and agile."} 150 | {"id": 432, "premise": "She must have been unfit.", "hypothesis": "She must have been like a snow leopard.", "label": "Contradiction", "explanation": "Snow leopards are known for their grace, power, and agility, so to say she was like one would imply that she was in good shape and not unfit"} 151 | {"id": 439, "premise": "Ali would never be caught in something garish.", "hypothesis": "Ali never would have worn something like a new jersey devils t-shirt.", "label": "Entailment", "explanation": "The New Jersey Devils are a hockey team with a red and black color scheme, which is considered garish by many people."} 152 | {"id": 439, "premise": "Ali would never be caught in something tasteful.", "hypothesis": "Ali never would have worn something like a new jersey devils t-shirt.", "label": "Contradiction", "explanation": "A New Jersey Devils t-shirt is a hockey team shirt, which are generally considered to be tasteless"} 153 | {"id": 440, "premise": "The click of the padlock echoed loudly in the silence and stillness.", "hypothesis": "In the silence and the stillness, the click of the padlock seemed like a rifle report.", "label": "Entailment", "explanation": "A rifle report is a loud, sudden noise, so the simile is emphasizing how loud and sudden the padlock click was."} 154 | {"id": 440, "premise": "The click of the padlock was muffled in the silence and stillness.", "hypothesis": "In the silence and the stillness, the click of the padlock seemed like a rifle report.", "label": "Contradiction", "explanation": "A rifle report is a loud noise, whereas a muffled noise is soft, so the two cannot be compared"} 155 | {"id": 446, "premise": "Even though he was strong in his magic, it felt weak compared to the inferno I'd faced when I killed Mab.", "hypothesis": "He was strong in his magic but, compared to the blazing inferno that i'd faced when i'd killed mab, his power felt like a candle flame.", "label": "Entailment", "explanation": "A candle flame is small and weak compared to an inferno, so he feels weak."} 156 | {"id": 446, "premise": "Even though he was strong in his magic, it felt powerful compared to the inferno I'd faced when I killed Mab.", "hypothesis": "He was strong in his magic but, compared to the blazing inferno that i'd faced when i'd killed mab, his power felt like a candle flame.", "label": "Contradiction", "explanation": "A candle flame is a weak light, it would not be considered powerful"} 157 | {"id": 453, "premise": "Her emotions were wild.", "hypothesis": "Her emotions spun like a roulette wheel.", "label": "Entailment", "explanation": "A roulette wheel is a device that is used for gambling, and it is known for being wild and unpredictable, so her emotions are wild."} 158 | {"id": 453, "premise": "Her emotions were stable.", "hypothesis": "Her emotions spun like a roulette wheel.", "label": "Contradiction", "explanation": "A roulette wheel is constantly spinning and changing, so her emotions being like a roulette wheel would imply that they were constantly changing and not stable"} 159 | {"id": 457, "premise": "Shumenko was standing on a catwalk four yards in the air, leaning over a large apparatus made of stainless steel.", "hypothesis": "Shumenko stood on a catwalk four yards in the air, bent over a stainless-steel apparatus like the size of a blue whale.", "label": "Entailment", "explanation": "The size of a blue whale is very large, indicating that the apparatus is also very large."} 160 | {"id": 457, "premise": "Shumenko was standing on a catwalk four yards in the air, leaning over a small apparatus made of stainless steel.", "hypothesis": "Shumenko stood on a catwalk four yards in the air, bent over a stainless-steel apparatus like the size of a blue whale.", "label": "Contradiction", "explanation": "The blue whale is the largest mammal in the world, so saying the apparatus was the size of a blue whale would make it very large, not small"} 161 | {"id": 467, "premise": "He ran fast.", "hypothesis": "He ran like an olympic sprinter.", "label": "Entailment", "explanation": "Olympic sprinters are some of the fastest runners in the world, meaning he ran very fast."} 162 | {"id": 467, "premise": "He ran slowly.", "hypothesis": "He ran like an olympic sprinter.", "label": "Contradiction", "explanation": "An Olympic sprinter is someone who runs extremely fast, so saying someone ran like one would imply they ran quickly, not slowly"} 163 | {"id": 468, "premise": "Smith pointed to the creature in the pin stripe suit that now had a deep hole in its head.", "hypothesis": "Smith pointed to the creature in the pin stripe suit that now had a hole in its head like a crater.", "label": "Entailment", "explanation": "A crater is a deep hole, so the simile is emphasizing the depth of the hole in the creature's head."} 164 | {"id": 468, "premise": "Smith pointed to the creature in the pin stripe suit that now had a shallow hole in its head.", "hypothesis": "Smith pointed to the creature in the pin stripe suit that now had a hole in its head like a crater.", "label": "Contradiction", "explanation": "A crater is a deep, bowl-shaped hole, so saying the hole is like a crater would imply it is deep"} 165 | {"id": 485, "premise": "I made quick work of one with my knife and slipped inside undetected", "hypothesis": "I made quick work of one with my knife and slipped inside like a night shadow.", "label": "Entailment", "explanation": "A night shadow is undetected because it is dark, so the person in the simile sentence is saying they slipped inside undetected."} 166 | {"id": 485, "premise": "I made quick work of one with my knife and slipped inside clumsily", "hypothesis": "I made quick work of one with my knife and slipped inside like a night shadow.", "label": "Contradiction", "explanation": "Night shadows are stealthy, so slipping in like a shadow would be done without making noise or being clumsy"} 167 | {"id": 490, "premise": "When you were younger, your creativity was unmatched.", "hypothesis": "When you were but a child, your creative power was like an lesson to students and teachers alike.", "label": "Entailment", "explanation": "A lesson is a period of teaching where students learn from the teacher, so saying it was a lesson for student and teacher alike means it was very superior and unmatched."} 168 | {"id": 490, "premise": "When you were younger, your creativity was average.", "hypothesis": "When you were but a child, your creative power was like an lesson to students and teachers alike.", "label": "Contradiction", "explanation": "A teacher is someone who is knowledgeable and skilled in a particular subject, and is hired to share that knowledge with students, so if creativity was a lesson to the teacher themselves it was not average but extraordinary"} 169 | {"id": 493, "premise": "Netflix & Chill just went horribly wrong", "hypothesis": "Netflix & Chill just went like a glitching code.", "label": "Entailment", "explanation": "A glitching code is a code that is not working properly, so Netflix & Chill went wrong in the same way."} 170 | {"id": 493, "premise": "Netflix & Chill just went great", "hypothesis": "Netflix & Chill just went like a glitching code.", "label": "Contradiction", "explanation": "A glitching code would imply that there were errors and the program was not running smoothly, whereas saying it went great would imply that it ran smoothly with no errors"} 171 | {"id": 494, "premise": "If the wrong person gets ahold of it, it would be catastrophic.", "hypothesis": "If it falls into the wrong hands it would be like the start of World War III.", "label": "Entailment", "explanation": "World War III would be catastrophic, so if it falls into the wrong hands, it would be just as bad."} 172 | {"id": 494, "premise": "If the wrong person gets ahold of it, it would be beneficial.", "hypothesis": "If it falls into the wrong hands it would be like the start of World War III.", "label": "Contradiction", "explanation": "World War III would bring devastation and catastrophy to humankind, so it cannot be described as beneficial"} 173 | {"id": 497, "premise": "You realize that everything that happened was intentional after taking the pill.", "hypothesis": "After taking a pill, you suddenly realize that it all was like predestination.", "label": "Entailment", "explanation": "Predestination is the belief that everything that happens is part of a larger plan."} 174 | {"id": 497, "premise": "You realize that everything that happened was accidental after taking the pill.", "hypothesis": "After taking a pill, you suddenly realize that it all was like predestination.", "label": "Contradiction", "explanation": "Predestination is the belief that your life has been determined in advance and that you cannot change it, whereas if everything was accidental, it would mean that your life is not determined in advance and that you can change it"} 175 | {"id": 499, "premise": "Your dog is very excited to see you and is acting quite peculiar.", "hypothesis": "Your dog is very excited to see you and is acting like a hyperactive sloth.", "label": "Entailment", "explanation": "Sloths are known for their peculiar behavior, which in this case is being very excited and hyperactive."} 176 | {"id": 499, "premise": "Your dog is very excited to see you and is acting quite normal.", "hypothesis": "Your dog is very excited to see you and is acting like a hyperactive sloth.", "label": "Contradiction", "explanation": "A sloth is a slow-moving mammal, so a loth acting hyperactively would be very weird"} 177 | {"id": 500, "premise": "You wake up to find that is unusually quiet", "hypothesis": "You wake up to find that is like a hibernating forest.", "label": "Entailment", "explanation": "A hibernating forest is a place where animals are dormant and there is little to no noise, so the context refers to waking up and finding a quiet atmosphere."} 178 | {"id": 500, "premise": "You wake up to find that it is unusually loud.", "hypothesis": "You wake up to find that is like a hibernating forest.", "label": "Contradiction", "explanation": "Forests are typically very quiet, so it would be impossible for it to be loud if it is like a hibernating forest"} 179 | {"id": 506, "premise": "They said that constructing the flat pack TV cabinet would be easy.", "hypothesis": "Constructing the flat pack TV cabinet was meant to be like turning on a light.", "label": "Entailment", "explanation": "Turning on a light is easy, so they are saying that constructing the flat pack TV cabinet would be easy."} 180 | {"id": 506, "premise": "They said that constructing the flat pack TV cabinet would be difficult.", "hypothesis": "Constructing the flat pack TV cabinet was meant to be like turning on a light.", "label": "Contradiction", "explanation": "Something that is easy to do is often described as being like turning on a light, whereas something that is difficult to do would be the opposite"} 181 | {"id": 513, "premise": "This story, however, is about a god who is as incompetent as he is useless.", "hypothesis": "This tale, however, speaks of a god who's as incompetent as he is like a gnat.", "label": "Entailment", "explanation": "A gnat is a small, insignificant insect, so to say that someone is as incompetent as a gnat would be to say that they are very incompetent."} 182 | {"id": 513, "premise": "This story, however, is about a god who is as incompetent as he is useful.", "hypothesis": "This tale, however, speaks of a god who's as incompetent as he is like a gnat.", "label": "Contradiction", "explanation": "A gnat is a very small and insignificant insect, while being like a gnat would imply being small and insignificant"} 183 | {"id": 528, "premise": "You wouldn't notice anything different about them until you see their green eyes.", "hypothesis": "Until you notice their eyes are glowing like a black cat's.", "label": "Entailment", "explanation": "Black cat's eyes are usually green which would glow more noticeably given their dark complexion"} 184 | {"id": 528, "premise": "You wouldn't notice anything different about them until you see their brown eyes.", "hypothesis": "Until you notice their eyes are glowing like a black cat's.", "label": "Contradiction", "explanation": "Black cat's eyes are usually green and not brown which would glow more noticeably given their dark complexion"} 185 | {"id": 538, "premise": "Even though they were visible to everyone, but were still inscrutable.", "hypothesis": "They were visible to all, but like blurred watercolored lines.", "label": "Entailment", "explanation": "The lines in a watercolor painting are usually blurred and not well-defined, so the simile is suggesting that the people are visible but not well-defined or understood."} 186 | {"id": 538, "premise": "Even though they were visible to everyone, they were still clear and comprehensible.", "hypothesis": "They were visible to all, but like blurred watercolored lines.", "label": "Contradiction", "explanation": "Blurred lines are not clear and comprehensible"} 187 | {"id": 542, "premise": "Clare felt each stirke painfully", "hypothesis": "Clare felt each stirke like it were the plunge of a poison-soaked blade.", "label": "Entailment", "explanation": "A poison-soaked blade would obviously cause pain, hence the simile is used to describe how Clare felt each strike."} 188 | {"id": 542, "premise": "Clare felt each stroke pleasurably.", "hypothesis": "Clare felt each stirke like it were the plunge of a poison-soaked blade.", "label": "Contradiction", "explanation": "A poison-soaked blade would cause pain, not pleasure"} 189 | {"id": 551, "premise": "Though the distance I have to cover is short, it feels brutal", "hypothesis": "Though the distance I have to cover is short, it feels like a scorching trek through Death Valley.", "label": "Entailment", "explanation": "Death Valley is a place known for its heat, so it would be brutal to trek through."} 190 | {"id": 551, "premise": "Though the distance I have to cover is short, it feels gentle.", "hypothesis": "Though the distance I have to cover is short, it feels like a scorching trek through Death Valley.", "label": "Contradiction", "explanation": "Death Valley is a very hot desert, with temperatures reaching up to 50 degrees Celsius, so saying the trek is like a scorching trek through Death Valley implies it is very hot, and not gentle"} 191 | {"id": 570, "premise": "The basement they shut him in looked dangerously claustrophobic.", "hypothesis": "They had shut him in a basement that looked like a freight elevator.", "label": "Entailment", "explanation": "A freight elevator is typically small and cramped, not safe and spacious and is often perceived to be claustrophobic"} 192 | {"id": 570, "premise": "The basement they shut him in looked safe and spacious.", "hypothesis": "They had shut him in a basement that looked like a freight elevator.", "label": "Contradiction", "explanation": "A freight elevator is typically small and cramped, not safe and spacious"} 193 | {"id": 572, "premise": "The city was beautiful", "hypothesis": "The city was like a well-cut diamond.", "label": "Entailment", "explanation": "A well-cut diamond is a thing of beauty, so the city being like a well-cut diamond would imply that it, too, is beautiful."} 194 | {"id": 572, "premise": "The city was ugly.", "hypothesis": "The city was like a well-cut diamond.", "label": "Contradiction", "explanation": "A well-cut diamond is beautiful and has symmetry, whereas an ugly city would not have these qualities"} 195 | {"id": 581, "premise": "We have learned that the Earth's ending is near and unavoidable.", "hypothesis": "We have learned that the ending of the Earth is like a blak hole's event horizon.", "label": "Entailment", "explanation": "A black hole's event horizon is the point of no return, past which anything that enters is drawn in irrevocably."} 196 | {"id": 581, "premise": "We have learned that the Earth's ending is distant and preventable.", "hypothesis": "We have learned that the ending of the Earth is like a blak hole's event horizon.", "label": "Contradiction", "explanation": "A black hole's event horizon is the point of no return, after which anything that enters is sucked in and can never escape, so saying the Earth's ending is like a black hole's event horizon implies that it is inevitable and not preventable"} 197 | {"id": 599, "premise": "It was red", "hypothesis": "It was like a blooming rose.", "label": "Entailment", "explanation": "A blooming rose is red."} 198 | {"id": 599, "premise": "It was blue", "hypothesis": "It was like a blooming rose.", "label": "Contradiction", "explanation": "A blooming rose is typically red, so saying it was blue contradicts the simile"} 199 | {"id": 604, "premise": "Even though the suspect is squeaky clean, you know they're guilty.", "hypothesis": "The suspect on your biggest case is squeaky clean, but you know it's like the blood is on their hands.", "label": "Entailment", "explanation": "The simile is saying that even though the suspect appears to be clean, they are actually guilty."} 200 | {"id": 604, "premise": "Even though the suspect is squeaky clean, you know they're innocent.", "hypothesis": "The suspect on your biggest case is squeaky clean, but you know it's like the blood is on their hands.", "label": "Contradiction", "explanation": "The phrase \"like the blood is on their hands\" means that they are guilty, even if there is no evidence to prove it"} 201 | {"id": 605, "premise": "Dreams are terrible when life is going well.", "hypothesis": "When times are good, your dreams are like a manifestation of your worst anxieties.", "label": "Entailment", "explanation": "Dreams are usually a manifestation of anxiety or fear, so when life is going well and there are no anxieties, the dreams are worst because there is nothing to fear."} 202 | {"id": 605, "premise": "Dreams are wonderful when life is going well.", "hypothesis": "When times are good, your dreams are like a manifestation of your worst anxieties.", "label": "Contradiction", "explanation": "A manifestation is a physical embodiment or realization of something, so if your dreams are like a manifestation of your worst anxieties, it means they are physical embodiments of your anxieties, which would not be wonderful"} 203 | {"id": 618, "premise": "he wasn't scared", "hypothesis": "He was as frightened as a bear in the woods", "label": "Entailment", "explanation": "A bear in the woods is one of the strongest animals, so it would itself be frightening, and typically would not feel scared at all"} 204 | {"id": 618, "premise": "he was scared", "hypothesis": "He was as frightened as a bear in the woods", "label": "Contradiction", "explanation": "A bear in the woods is one of the strongest animals, so it would itself be frightening, and typically would not feel scared at all"} 205 | {"id": 620, "premise": "Their yard is ugly and unkempt.", "hypothesis": "Their yard is as pretty as a dunghill.", "label": "Entailment", "explanation": "A dunghill is an unsightly pile of manure, so the simile is saying that their yard is just unkempt ,ugly and unkempt"} 206 | {"id": 620, "premise": "Their yard is beautiful.", "hypothesis": "Their yard is as pretty as a dunghill.", "label": "Contradiction", "explanation": "A dunghill is a pile of manure, which is not pretty, so saying that their yard is beautiful would be contradictory"} 207 | {"id": 622, "premise": "The watermelon was sour", "hypothesis": "The watermelon was as sweet as a lemon", "label": "Entailment", "explanation": "A lemon is a sour fruit, so the watermelon must be sour if it is as sweet as a lemon."} 208 | {"id": 622, "premise": "The watermelon was sweet", "hypothesis": "The watermelon was as sweet as a lemon", "label": "Contradiction", "explanation": "Lemons are sour, so the simile is saying that the watermelon is not as sweet as one would expect"} 209 | {"id": 630, "premise": "The future seems dull", "hypothesis": "The future seems as bright as the Arctic winter", "label": "Entailment", "explanation": "The Arctic winter is very dark and bleak, so the future seeming as bright as the Arctic winter would mean that it is not bright at all."} 210 | {"id": 630, "premise": "The future seems bright", "hypothesis": "The future seems as bright as the Arctic winter", "label": "Contradiction", "explanation": "The Arctic winter is the darkest time of the year, so saying the future is as bright as the Arctic winter would mean that the future is very dark"} 211 | {"id": 641, "premise": "My skin is coarse and rough", "hypothesis": "My skin is as smooth as gravel", "label": "Entailment", "explanation": "Gravel is a type of stone that is very rough, so if someone's skin is as smooth as gravel, that means their skin is very coarse and rough."} 212 | {"id": 641, "premise": "My skin is smooth", "hypothesis": "My skin is as smooth as gravel", "label": "Contradiction", "explanation": "Gravel is a rough material, so saying skin is as smooth as gravel would mean that the skin is not actually smooth"} 213 | {"id": 649, "premise": "The wife was a horrible cook", "hypothesis": "The wife was as good of a cook as a child", "label": "Entailment", "explanation": "A child is not a good cook, therefore the wife must be a bad cook if she is as good as a child."} 214 | {"id": 649, "premise": "The wife was a great cook", "hypothesis": "The wife was as good of a cook as a child", "label": "Contradiction", "explanation": "A child is not typically known for being a great cook, so the wife being as good of a cook as a child would imply that she is not a great cook"} 215 | {"id": 650, "premise": "Your face is distorted and strange", "hypothesis": "Your face is as handsome as a painting by Picasso", "label": "Entailment", "explanation": "A Picasso painting is often distorted and strange, hence the simile."} 216 | {"id": 650, "premise": "Your face is handsome", "hypothesis": "Your face is as handsome as a painting by Picasso", "label": "Contradiction", "explanation": "A painting by Picasso is full of abstract shapes and colors, which may not be traditionally considered handsome"} 217 | {"id": 655, "premise": "She did not keep still", "hypothesis": "She kept as still as a bouncing superball", "label": "Entailment", "explanation": "A superball is a ball that is constantly moving and bouncing, so the simile is saying that she was not still at all."} 218 | {"id": 655, "premise": "She kept still", "hypothesis": "She kept as still as a bouncing superball", "label": "Contradiction", "explanation": "A superball is a ball that is very bouncy, so it is always in motion, whereas the phrase \"kept still\" means she was not moving"} 219 | {"id": 668, "premise": "The shirt was ugly.", "hypothesis": "The shirt was as cute as mud.", "label": "Entailment", "explanation": "Mud\" is often used to describe something that is dirty or unpleasant, hence the simile is indicating that the shirt is not attractive."} 220 | {"id": 668, "premise": "The shirt was cute.", "hypothesis": "The shirt was as cute as mud.", "label": "Contradiction", "explanation": "The phrase \"as cute as mud\" is used to describe something that is not cute at all, so the simile implies that the shirt is actually not cute"} 221 | {"id": 677, "premise": "The woman was cruel", "hypothesis": "The woman was as kind as a prisoner", "label": "Entailment", "explanation": "A prisoner is someone who is confined for a crime, so a kind prisoner would be a contradiction in terms and the woman was cruel actually."} 222 | {"id": 677, "premise": "The woman was kind", "hypothesis": "The woman was as kind as a prisoner", "label": "Contradiction", "explanation": "A prisoner is someone who is incarcerated, and most likely is not kind because they committed some bad crime, therefore the simile is saying that the woman is very cruel"} 223 | {"id": 680, "premise": "The story was fun and happy.", "hypothesis": "The story was as sad as a party", "label": "Entailment", "explanation": "This simile is incorrect because a party is generally a happy event, while a sad story is the opposite."} 224 | {"id": 680, "premise": "The story was sad", "hypothesis": "The story was as sad as a party", "label": "Contradiction", "explanation": "A party is an entertaining and fun event, which is the opposite of sad"} 225 | {"id": 682, "premise": "The student is dumb.", "hypothesis": "This student is as clever as a rock", "label": "Entailment", "explanation": "A rock is not a clever object, therefore the student must not be clever either."} 226 | {"id": 682, "premise": "The student is smart.", "hypothesis": "This student is as clever as a rock", "label": "Contradiction", "explanation": "A rock does not have the capacity for thought or cleverness, therefore saying someone is as clever as a rock would be implying that the person is not clever"} 227 | {"id": 688, "premise": "This field is not grassy.", "hypothesis": "This field is as grassy as the Sahara desert.", "label": "Entailment", "explanation": "The Sahara desert is a very dry, arid place with very little vegetation."} 228 | {"id": 688, "premise": "This field is very grassy.", "hypothesis": "This field is as grassy as the Sahara desert.", "label": "Contradiction", "explanation": "The Sahara desert is a large desert that is very dry and has no grass, so saying something is as grassy as the Sahara desert means it is not very grassy"} 229 | {"id": 696, "premise": "The baby hardly cried", "hypothesis": "the baby cried as hard as a butterfly flapping a broken wing", "label": "Entailment", "explanation": "A butterfly flapping a broken wing would not be able to fly, hence it would not be able to flap its wings very hard."} 230 | {"id": 696, "premise": "The baby cried hard", "hypothesis": "the baby cried as hard as a butterfly flapping a broken wing", "label": "Contradiction", "explanation": "A butterfly flapping a broken wing is not moving very fast or with much force, so the simile is saying that the baby wasn't crying very hard"} 231 | {"id": 703, "premise": "My dinner was cold", "hypothesis": "My dinner was as hot as snow", "label": "Entailment", "explanation": "Snow is cold so this simile is used to emphasize how cold the dinner was."} 232 | {"id": 703, "premise": "My dinner was hot", "hypothesis": "My dinner was as hot as snow", "label": "Contradiction", "explanation": "Snow is cold, so saying that something is as hot as snow means that it is very cold"} 233 | {"id": 707, "premise": "The employee is dumb.", "hypothesis": "The employee is sharp as a bowling ball", "label": "Entailment", "explanation": "A bowling ball is a heavy, round object that is used to knock down pins, it does not have any sharp corners, hence the person is not sharp which they are dumb"} 234 | {"id": 707, "premise": "The employee is smart.", "hypothesis": "The employee is sharp as a bowling ball", "label": "Contradiction", "explanation": "A bowling ball is a heavy, round object that is used to knock down pins, it does not have any sharp corners, hence the person is not sharp which means they are not smart"} 235 | {"id": 708, "premise": "The girl's personality wasn't bitter at all.", "hypothesis": "The girl's personality was as bitter as vanilla ice cream", "label": "Entailment", "explanation": "Vanilla icecream is a very sweet flavor, so the simile is saying that the girl's personality is the opposite of bitter."} 236 | {"id": 708, "premise": "The girl's personality was very bitter.", "hypothesis": "The girl's personality was as bitter as vanilla ice cream", "label": "Contradiction", "explanation": "Vanilla is a sweet flavor, so saying that the girl's personality was as bitter as vanilla ice cream would mean that it was not very bitter"} 237 | {"id": 709, "premise": "This item is not at all essential.", "hypothesis": "This is as essential as diamonds", "label": "Entailment", "explanation": "Diamonds are a luxury item and are not at all essential so the simile is saying that the item is not at all essential"} 238 | {"id": 709, "premise": "This item is very essential.", "hypothesis": "This is as essential as diamonds", "label": "Contradiction", "explanation": "Diamonds are a luxury item and are not at all essential"} 239 | {"id": 716, "premise": "Their marriage is exciting and unexpected", "hypothesis": "Their marriage is as drab as an lsd trip", "label": "Entailment", "explanation": "An LSD trip is usually characterized by colorful visuals and intense emotions, so a marriage that is as drab as an LSD trip would be quite exciting actually in comparison."} 240 | {"id": 716, "premise": "Their marriage is boring", "hypothesis": "Their marriage is as drab as an lsd trip", "label": "Contradiction", "explanation": "An LSD trip is a psychedelic experience that is usually colorful and stimulating, hence saying it is boring would be contradictory"} 241 | {"id": 727, "premise": "The coffee was very light", "hypothesis": "the coffee was as dark as snow in a field of cotton", "label": "Entailment", "explanation": "The coffee was very light in color, like snow in a field of cotton because both are white in color and would be not ver prominent."} 242 | {"id": 727, "premise": "The coffee was very dark", "hypothesis": "the coffee was as dark as snow in a field of cotton", "label": "Contradiction", "explanation": "Snow in a field of cotton would be one of the whitest things, so saying the coffee is as dark as that is emphasizing how light it is"} 243 | {"id": 736, "premise": "It would be a brief journey", "hypothesis": "The journey would be as long as a commercial", "label": "Entailment", "explanation": "A commercial is a very short advertisement, so the journey would be very brief."} 244 | {"id": 736, "premise": "The journey would be long", "hypothesis": "The journey would be as long as a commercial", "label": "Contradiction", "explanation": "A commercial break is a short intermission in a television program, hence the journey would be short, not long"} 245 | {"id": 740, "premise": "That branch is weak", "hypothesis": "That branch is as strong as string", "label": "Entailment", "explanation": "String is not a strong material, therefore the branch must not be strong either."} 246 | {"id": 740, "premise": "That branch is strong", "hypothesis": "That branch is as strong as string", "label": "Contradiction", "explanation": "String is not a strong material, so saying that the branch is as strong as string means that the branch is not very strong"} 247 | {"id": 745, "premise": "The people weren't scared at all.", "hypothesis": "The people were as scared as a lion.", "label": "Entailment", "explanation": "A lion is a very large and dangerous animal, it would not be scared but rather scary itself"} 248 | {"id": 745, "premise": "The people were very scared.", "hypothesis": "The people were as scared as a lion.", "label": "Contradiction", "explanation": "A lion is a very large and dangerous animal, it would not be scared but rather scary itself"} 249 | {"id": 746, "premise": "He's not smart at all", "hypothesis": "He's smart as a brick", "label": "Entailment", "explanation": "A brick is an inanimate object, therefore not smart."} 250 | {"id": 746, "premise": "He's very smart", "hypothesis": "He's smart as a brick", "label": "Contradiction", "explanation": "A brick is an inanimate object, therefore it would not be considered smart"} -------------------------------------------------------------------------------- /train_I->OR.sh: -------------------------------------------------------------------------------- 1 | export BS=16; 2 | PYTHONPATH=../../../src 3 | USE_TF=0 4 | 5 | python ./run_translation.py \ 6 | --output_dir ./output \ 7 | --model_name_or_path t5-3b \ 8 | --cache_dir ./cache \ 9 | --evaluation_strategy epoch \ 10 | --save_strategy epoch \ 11 | --do_train \ 12 | --do_eval \ 13 | --train_file ./FLUTEfinaltrain.json \ 14 | --validation_file ./FLUTEfinalvaljson \ 15 | --overwrite_output_dir \ 16 | --max_source_length 1024 \ 17 | --max_target_length 256 \ 18 | --num_train_epochs 10 \ 19 | --gradient_accumulation_steps 64 \ 20 | --per_device_train_batch_size $BS \ 21 | --per_device_eval_batch_size $BS \ 22 | --source_lang en_XX \ 23 | --target_lang en_XX 24 | --------------------------------------------------------------------------------