├── .gitignore ├── ACL2018.pptx ├── LICENSE ├── ML_Comparisons.py ├── README.md ├── context2vec.py ├── data ├── 1x1_encode_map.pkl ├── 1x1_geonames.pkl ├── 1x1_outliers_map.pkl ├── 1x1_reverse_map.pkl ├── 2x2_encode_map.pkl ├── 2x2_geonames.pkl ├── 2x2_outliers_map.pkl ├── 2x2_reverse_map.pkl ├── GeoVirus.xml ├── WikToR.xml ├── eval_geovirus.txt ├── eval_geovirus_gold.txt ├── eval_lgl.txt ├── eval_lgl_gold.txt ├── eval_wiki.txt ├── eval_wiki_gold.txt.zip ├── geovirus.txt ├── geovirus_gold.txt ├── iaa_answers.txt ├── iaa_check.txt ├── iaa_test.txt ├── lgl.txt ├── lgl.xml ├── lgl_gold.txt ├── wiki.txt ├── wiki_gold.txt └── words2index.pkl ├── geoparse.py ├── geovirus.py ├── melbourne-augmenting-geocoding.pdf ├── preprocessing.py ├── subsample.py ├── test.py ├── text2mapVec.py └── train.py /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | .DS_Store 12 | env/ 13 | build/ 14 | develop-eggs/ 15 | dist/ 16 | downloads/ 17 | eggs/ 18 | .eggs/ 19 | lib/ 20 | lib64/ 21 | parts/ 22 | sdist/ 23 | var/ 24 | *.egg-info/ 25 | .installed.cfg 26 | *.egg 27 | .idea/ 28 | 29 | # PyInstaller 30 | # Usually these files are written by a python script from a template 31 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 32 | *.manifest 33 | *.spec 34 | 35 | # Installer logs 36 | pip-log.txt 37 | pip-delete-this-directory.txt 38 | 39 | # Unit test / coverage reports 40 | htmlcov/ 41 | .tox/ 42 | .coverage 43 | .coverage.* 44 | .cache 45 | nosetests.xml 46 | coverage.xml 47 | *,cover 48 | .hypothesis/ 49 | 50 | # Translations 51 | *.mo 52 | *.pot 53 | 54 | # Django stuff: 55 | *.log 56 | local_settings.py 57 | 58 | # Flask stuff: 59 | instance/ 60 | .webassets-cache 61 | 62 | # Scrapy stuff: 63 | .scrapy 64 | 65 | # Sphinx documentation 66 | docs/_build/ 67 | 68 | # PyBuilder 69 | target/ 70 | 71 | # IPython Notebook 72 | .ipynb_checkpoints 73 | 74 | # pyenv 75 | .python-version 76 | 77 | # celery beat schedule file 78 | celerybeat-schedule 79 | 80 | # dotenv 81 | .env 82 | 83 | # virtualenv 84 | venv/ 85 | ENV/ 86 | 87 | # Spyder project settings 88 | .spyderproject 89 | 90 | # Rope project settings 91 | .ropeproject 92 | 93 | # my other temporary files 94 | data/eval_wiki_gold.txt 95 | errors.tsv 96 | gpu/ 97 | output.txt -------------------------------------------------------------------------------- /ACL2018.pptx: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/milangritta/Geocoding-with-Map-Vector/69e6e590c56930ed80d346f6c2da6d58056182e3/ACL2018.pptx -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | {one line to give the program's name and a brief idea of what it does.} 635 | Copyright (C) {year} {name of author} 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | {project} Copyright (C) {year} {fullname} 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /ML_Comparisons.py: -------------------------------------------------------------------------------- 1 | import sqlite3 2 | import sys 3 | from geopy.distance import great_circle 4 | from sklearn.ensemble import RandomForestClassifier 5 | from sklearn.naive_bayes import MultinomialNB 6 | from preprocessing import generate_arrays_from_file_map2vec, index_to_coord, get_coordinates, generate_strings_from_file 7 | from preprocessing import REVERSE_MAP_2x2 8 | from preprocessing import print_stats 9 | import numpy as np 10 | from sklearn.externals import joblib 11 | 12 | # For command line use, type: python test.py such as lgl_gold or wiki (see file names) 13 | if len(sys.argv) > 1: 14 | data = sys.argv[1] 15 | else: 16 | data = u"lgl" 17 | 18 | X, Y = [], [] 19 | clf = MultinomialNB() 20 | classes = range(len(REVERSE_MAP_2x2)) 21 | # clf = RandomForestClassifier() 22 | for (x, y) in generate_arrays_from_file_map2vec(u"../data/train_wiki_uniform.txt", looping=False): 23 | X.extend(x[0]) 24 | Y.extend(np.argmax(y, axis=1)) 25 | # -------- Uncomment for Naive Bayes ------------- 26 | if len(X) > 25000: 27 | print(u"Training with:", len(X), u"examples.") 28 | clf.partial_fit(X, Y, classes) 29 | X, Y = [], [] 30 | # ------------------------------------------------ 31 | 32 | print(u"Training with:", len(X), u"examples.") 33 | clf.partial_fit(X, Y, classes) # Naive Bayes only! 34 | # clf.fit(X, Y) # Random Forest 35 | joblib.dump(clf, u'../data/bayes.pkl') # saves the model to file 36 | 37 | # ------------------------------------- END OF TRAINING, BEGINNING OF TESTING ----------------------------------- 38 | 39 | X = [] 40 | final_errors = [] 41 | clf = joblib.load(u'../data/bayes.pkl') 42 | test_file = u"data/eval_" + data + u".txt" # which data to test on? 43 | 44 | for (x, y) in generate_arrays_from_file_map2vec(test_file, looping=False): 45 | X.extend(x[0]) # Load test instances 46 | 47 | print(u"Testing with:", len(X), u"examples.") 48 | conn = sqlite3.connect(u'../data/geonames.db') 49 | 50 | for x, (y, name, context) in zip(clf.predict(X), generate_strings_from_file(test_file)): 51 | p = index_to_coord(REVERSE_MAP_2x2[x], 2) 52 | candidates = get_coordinates(conn.cursor(), name) 53 | 54 | if len(candidates) == 0: 55 | print(u"Don't have an entry for", name, u"in GeoNames") 56 | raise Exception(u"Check your database, buddy :-)") 57 | 58 | # candidates = [candidates[0]] # Uncomment for population heuristic. 59 | # THE ABOVE IS THE POPULATION ONLY BASELINE IMPLEMENTATION 60 | 61 | best_candidate = [] 62 | max_pop = candidates[0][2] 63 | bias = 0.9 # bias parameter, see 64 | for candidate in candidates: 65 | err = great_circle(p, (float(candidate[0]), float(candidate[1]))).km 66 | best_candidate.append((err - (err * max(1, candidate[2]) / max(1, max_pop)) * bias, (float(candidate[0]), float(candidate[1])))) 67 | best_candidate = sorted(best_candidate, key=lambda (a, b): a)[0] 68 | final_errors.append(great_circle(best_candidate[1], y).km) 69 | 70 | print_stats(final_errors) 71 | print(u"Done testing:", test_file) 72 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Which Melbourne? Augmenting Geocoding with Maps 2 | 3 | ### Resources accompanying the ACL 2018 long paper, presented in Melbourne, Australia. 4 | 5 | *The accepted pdf manuscript is also included in this directory (as is the .PPTX from the Melbourne presentation). The video recording of the Melbourne presentation can be found here (https://vimeo.com/285803462).* 6 | 7 | ##### Abstract 8 | The purpose of text geolocation is to associate geographic information contained 9 | in a document with a set (or sets) of coordinates, either implicitly by using linguistic 10 | features and/or explicitly by using geographic metadata combined with 11 | heuristics. We introduce a geocoder (location mention disambiguator) that achieves 12 | state-of-the-art (SOTA) results on three diverse datasets by exploiting the implicit 13 | lexical clues. Moreover, we propose a new method for systematic encoding of 14 | geographic metadata to generate two distinct views of the same text. To that end, 15 | we introduce the Map Vector (MapVec), a sparse representation obtained by plotting 16 | prior geographic probabilities, derived from population figures, on a World 17 | Map. We then integrate the implicit (language) and explicit (map) features to significantly 18 | improve a range of metrics. We also introduce an open-source dataset for 19 | geoparsing of news events covering global disease outbreaks and epidemics to help 20 | future evaluation in geoparsing. 21 | 22 | ##### Resources 23 | 24 | This repository contains the accompanying data and source code for CamCoder (toponym resolver) described in the paper. Additional data is required as the files are too large for GitHub, please download files from **https://www.repository.cam.ac.uk/handle/1810/277772**. 25 | 26 | #### Dependencies 27 | * Keras 2.2.0 https://keras.io/#installation 28 | * Tensorflow 1.8 https://www.tensorflow.org/install/ 29 | * Spacy 2.0 (also download a model https://spacy.io/usage/models) 30 | * Python 2.7+ and a recent version of sqlite, matplotlib, cpickle and geopy 31 | * The rest should be installed alongside the three major libraries 32 | * Next time I'll use Docker, too late now, sorry about that. 33 | 34 | ### Instructions 35 | * Download the `weights.zip` and `geonames.db.zip` files as a **minimum** (optional files available from *https://www.repository.cam.ac.uk/handle/1810/277772*). 36 | * Read the `README.txt` in the repository to learn about the contents. 37 | * Create a **data** folder *outside the root directory* to store the large files. N.B. There is already a data folder **inside** the root directory! This holds the small files. 38 | * Unzip the files into that directory, this will take up a few GBs of space. 39 | * For replication, use `test.py` and see further instructions in the code. That should run out of the box if you followed the previous instructions. If not, get in touch! 40 | * To tweak the model, use `train.py`, see comments inside the script for more info. 41 | 42 | Use a GPU, if you can, a CPU epoch takes such a looooooong time, it's only worth it for small jobs. Contact me on :envelope: *mg711 at cam dot ac dot uk* :envelope: if you need any help with reproduction or some other aspect of this work at any time. After graduation, find me on Twitter/milangritta or raise an issue/ticket. 43 | 44 | ### Tools 45 | I included a couple of 'tools' for applied scientists and tinkerers in case you want to parse your own text and/or want to compare system performance with your research. 46 | #### text2mapVec.py 47 | This is a simple function `buildMapVec(text)` that turns text into a **Map Vector** i.e. extracts locations/toponyms with **Spacy NER** and creates the 'bag of locations' or the Map Vector as an additional feature vector to be used in a downstream task. 48 | 49 | *NOTE: The speed of execution won't be a record breaker, this is research code, I'm really busy trying to finish the PhD, sorry, I don't have time to rewrite it from scratch using proper software engineering principles. I hope you understand. Feel free to fork and edit.* 50 | #### geoparse.py 51 | Unline most (maybe all) geoparsers, CamCoder can perform *geotagging* (NER) and *geocoding* separately. Use (1.) for the full pipeline and (2.) for toponym resolution only. 52 | 1. To geocode with NER: Use `geoparse(text)`, instructions in the code. 53 | 2. To geocode with Oracle: This will be slightly more laborious as you will need the `generate_evaluation_data(corpus, file_name)` function in `preprocessing.py`. First, save your evaluation dataset in the format of `data/lgl.txt` (name,,name,,lat,,lon,,start,end) then you don't have to modify any code. I think it's the best option. Once you have generated machine-readable data with that function, you're ready to `test.py` the performance. 54 | 55 | *NOTE: CamCoder uses **Spacy NER** for Named Entity Recognition. The reported F-Scores for each model can be found here https://spacy.io/models/en, not that great and will certainly affect performance. Use **Oracle NER** for a scientifically adequate comparison. Oracle means you extract the entities separately with perfect fidelity, then evaluate toponym recognition in isolation. Also feel free to plug in a custom **NER tagger**, the code is extendable and should be well documented. Famous last words :-)* 56 | -------------------------------------------------------------------------------- /context2vec.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | import codecs 3 | import numpy as np 4 | import cPickle 5 | from keras import Input 6 | from keras.callbacks import ModelCheckpoint, EarlyStopping 7 | from keras.engine import Model 8 | from keras.layers.merge import concatenate 9 | from keras.layers import Embedding, Dense, Dropout, LSTM 10 | from preprocessing import BATCH_SIZE, EMBEDDING_DIMENSION, CONTEXT_LENGTH, UNKNOWN, TARGET_LENGTH 11 | from preprocessing import generate_arrays_from_file_lstm, ENCODING_MAP_2x2, ENCODING_MAP_1x1 12 | from subprocess import check_output 13 | 14 | print(u"Embedding Dimension:", EMBEDDING_DIMENSION) 15 | print(u"Input length (each side):", CONTEXT_LENGTH) 16 | word_to_index = cPickle.load(open(u"data/words2index.pkl")) 17 | print(u"Vocabulary Size:", len(word_to_index)) 18 | 19 | vectors = {UNKNOWN: np.ones(EMBEDDING_DIMENSION), u'0': np.ones(EMBEDDING_DIMENSION)} 20 | for line in codecs.open(u"../data/glove.twitter." + str(EMBEDDING_DIMENSION) + u"d.txt", encoding=u"utf-8"): 21 | if line.strip() == "": 22 | continue 23 | t = line.split() 24 | vectors[t[0]] = [float(x) for x in t[1:]] 25 | print(u'Vectors...', len(vectors)) 26 | 27 | emb_weights = np.zeros((len(word_to_index), EMBEDDING_DIMENSION)) 28 | oov = 0 29 | for w in word_to_index: 30 | if w in vectors: 31 | emb_weights[word_to_index[w]] = vectors[w] 32 | else: 33 | emb_weights[word_to_index[w]] = np.random.normal(size=(EMBEDDING_DIMENSION,), scale=0.3) 34 | oov += 1 35 | 36 | emb_weights = np.array([emb_weights]) 37 | print(u'Done preparing vectors...') 38 | print(u"OOV (no vectors):", oov) 39 | # -------------------------------------------------------------------------------------------------------------------- 40 | print(u'Building model...') 41 | embeddings = Embedding(len(word_to_index), EMBEDDING_DIMENSION, input_length=CONTEXT_LENGTH, weights=emb_weights) 42 | # shared embeddings between all language input layers 43 | 44 | forward = Input(shape=(CONTEXT_LENGTH,)) 45 | cwf = embeddings(forward) 46 | cwf = LSTM(300)(cwf) 47 | cwf = Dense(300)(cwf) 48 | cwf = Dropout(0.5)(cwf) 49 | 50 | backward = Input(shape=(CONTEXT_LENGTH,)) 51 | cwb = embeddings(backward) 52 | cwb = LSTM(300, go_backwards=True)(cwb) 53 | cwb = Dense(300)(cwb) 54 | cwb = Dropout(0.5)(cwb) 55 | 56 | # Uncomment this block for MAPVEC + CONTEXT2VEC model, also uncomment 2 lines further down, thanks! 57 | # You also need to uncomment a few lines in preprocessing.py, generate_arrays_from_file_lstm() function 58 | # mapvec = Input(shape=(len(ENCODING_MAP_1x1),)) 59 | # l2v = Dense(5000, activation='relu', input_dim=len(ENCODING_MAP_1x1))(mapvec) 60 | # l2v = Dense(1000, activation='relu')(l2v) 61 | # l2v = Dropout(0.5)(l2v) 62 | 63 | target_string = Input(shape=(TARGET_LENGTH,)) 64 | ts = Embedding(len(word_to_index), EMBEDDING_DIMENSION, input_length=TARGET_LENGTH, weights=emb_weights)(target_string) 65 | ts = LSTM(50)(ts) 66 | ts = Dense(50)(ts) 67 | ts = Dropout(0.5)(ts) 68 | 69 | inp = concatenate([cwf, cwb, ts]) 70 | # inp = concatenate([cwf, cwb, mapvec, ts]) # Uncomment for MAPVEC + CONTEXT2VEC 71 | inp = Dense(units=len(ENCODING_MAP_2x2), activation=u'softmax')(inp) 72 | model = Model(inputs=[forward, backward, target_string], outputs=[inp]) 73 | # model = Model(inputs=[forward, backward, mapvec, target_string], outputs=[inp]) # Uncomment for MAPVEC + CONTEXT2VEC 74 | model.compile(loss=u'categorical_crossentropy', optimizer=u'rmsprop', metrics=[u'accuracy']) 75 | 76 | print(u'Finished building model...') 77 | # -------------------------------------------------------------------------------------------------------------------- 78 | checkpoint = ModelCheckpoint(filepath=u"../data/weights.{epoch:02d}-{acc:.2f}.hdf5", verbose=0) 79 | early_stop = EarlyStopping(monitor=u'acc', patience=5) 80 | file_name = u"../data/train_wiki_uniform.txt" 81 | model.fit_generator(generate_arrays_from_file_lstm(file_name, word_to_index), 82 | steps_per_epoch=int(check_output(["wc", file_name]).split()[0]) / BATCH_SIZE, 83 | epochs=250, callbacks=[checkpoint, early_stop]) 84 | -------------------------------------------------------------------------------- /data/eval_wiki_gold.txt.zip: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/milangritta/Geocoding-with-Map-Vector/69e6e590c56930ed80d346f6c2da6d58056182e3/data/eval_wiki_gold.txt.zip -------------------------------------------------------------------------------- /data/geovirus.txt: -------------------------------------------------------------------------------- 1 | 2 | Sudan,,Sudan,,15,,32,,994,,999|| 3 | 4 | Texas,,Texas,,31,,-100,,86,,91||Africa,,Africa,,7.18,,21.09,,441,,447|| 5 | Nigeria,,Nigeria,,8,,10,,423,,430|| 6 | London,,London,,51.50,,-0.12,,28,,34||Sierra Leone,,Sierra Leone,,8.5,,-11.5,,342,,354|| 7 | 8 | Taiwan,,Taiwan,,25.03,,121.63,,46,,52||Taoyuan,,Taoyuan,,24.98,,121.31,,96,,103||Taiwan,,Taiwan,,25.03,,121.63,,253,,259|| 9 | 10 | Crawford County,,Crawford County,,41.68,,-80.11,,222,,237||Pennsylvania,,Pennsylvania,,41,,-77.5,,249,,261||Pennsylvania,,Pennsylvania,,41,,-77.5,,347,,359||Asia,,Asia,,29.84,,89.29,,505,,509||Europe,,Europe,,51,,17.5,,511,,517||Africa,,Africa,,7.18,,21.09,,522,,528||Pennsylvania,,Pennsylvania,,41,,-77.5,,934,,946|| 11 | 12 | Sweden,,Sweden,,63,,16,,100,,106||Baltic Sea,,Baltic Sea,,58,,20,,669,,679|| 13 | Egypt,,Egypt,,26,,30,,383,,388||Cairo,,Cairo,,30.04,,31.23,,398,,403|| 14 | United Kingdom,,United Kingdom,,55,,-3,,22,,36||Suffolk,,Suffolk,,52.16,,1,,256,,263||Norfolk,,Norfolk,,52.66,,1,,596,,603|| 15 | Hong Kong,,Hong Kong,,22.3,,114.2,,357,,366||Indonesia,,Indonesia,,-5,,120,,433,,442|| 16 | Jakarta,,Jakarta,,-6.2,,106.81,,298,,305|| 17 | England,,England,,51.5,,-0.11,,669,,676|| 18 | 19 | Russia,,Russia,,60,,90,,83,,89||Romania,,Romania,,46,,25,,91,,98||Macedonia,,Macedonia,,41.6,,21.7,,104,,113||China,,China,,35,,103,,202,,207||Hohhot,,Hohhot,,40.81,,111.65,,273,,279||Russia,,Russia,,60,,90,,368,,374||Moscow,,Moscow,,55.75,,37.61,,470,,476||Russia,,Russia,,60,,90,,587,,593||Macedonia,,Macedonia,,41.6,,21.7,,623,,632||Danube Delta,,Danube Delta,,45.33,,29.5,,790,,802|| 20 | Alaska,,Alaska,,64,,-150,,334,,340||Asia,,Asia,,29.84,,89.29,,379,,383||Europe,,Europe,,51,,17.5,,437,,443||Africa,,Africa,,7.18,,21.09,,448,,454||North America,,North America,,54.77,,-105.64,,553,,566|| 21 | Europe,,Europe,,51,,17.5,,20,,26||Greece,,Greece,,39,,22,,277,,283||Bulgaria,,Bulgaria,,42.75,,25.5,,285,,293||Italy,,Italy,,43,,12,,295,,300||Austria,,Austria,,47.33,,13.33,,302,,309||Germany,,Germany,,51,,9,,314,,321||Slovenia,,Slovenia,,46.11,,14.81,,430,,438||Croatia,,Croatia,,45.16,,15.5,,440,,447||Denmark,,Denmark,,56,,10,,452,,459||Africa,,Africa,,7.18,,21.09,,526,,532||Europe,,Europe,,51,,17.5,,536,,542||Russia,,Russia,,60,,90,,640,,646|| 22 | Egypt,,Egypt,,26,,30,,12,,17||Egypt,,Egypt,,26,,30,,159,,164|| 23 | Pyongyang,,Pyongyang,,39.01,,125.73,,94,,103||North Korea,,North Korea,,40,,127,,449,,460|| 24 | India,,India,,21,,78,,95,,100|| 25 | Iraq,,Iraq,,33,,44,,32,,36||United States,,United States,,40,,-100,,160,,173||Cairo,,Cairo,,30.04,,31.23,,197,,202||Iraq,,Iraq,,33,,44,,587,,591||Turkey,,Turkey,,39,,35,,636,,642|| 26 | 27 | Poland,,Poland,,52,,20,,18,,24||Pulawy,,Pulawy,,51.41,,21.96,,481,,487|| 28 | France,,France,,47,,2,,157,,163||France,,France,,47,,2,,165,,171||Austria,,Austria,,47.33,,13.33,,249,,256||Germany,,Germany,,51,,9,,258,,265||Greece,,Greece,,39,,22,,287,,293||Italy,,Italy,,43,,12,,298,,303||Croatia,,Croatia,,45.16,,15.5,,332,,339||Denmark,,Denmark,,56,,10,,344,,351||Europe,,Europe,,51,,17.5,,504,,510||France,,France,,47,,2,,816,,822|| 29 | United Kingdom,,United Kingdom,,55,,-3,,25,,39||Cellardyke,,Cellardyke,,56.21,,-2.7,,387,,397||Suffolk,,Suffolk,,52.16,,1,,473,,480|| 30 | Medan,,Medan,,3.58,,98.66,,371,,376|| 31 | China,,China,,35,,103,,53,,58|| 32 | Germany,,Germany,,51,,9,,95,,102||France,,France,,47,,2,,171,,177||France,,France,,47,,2,,421,,427||Europe,,Europe,,51,,17.5,,774,,780|| 33 | Romania,,Romania,,46,,25,,92,,99||Ciocile,,Ciocile,,44.81,,27.23,,277,,284||Russia,,Russia,,60,,90,,993,,999|| 34 | Michigan,,Michigan,,44,,-85,,99,,107||Lake Erie,,Lake Erie,,42.2,,-81.2,,124,,133||Monroe County,,Monroe County,,41.92,,-83.5,,171,,184||North America,,North America,,54.77,,-105.64,,575,,588||Asia,,Asia,,29.84,,89.29,,676,,680||Michigan,,Michigan,,44,,-85,,922,,930|| 35 | 36 | Dogubayazit,,Dogubayazit,,39.54,,44.08,,62,,73|| 37 | Illinois,,Illinois,,40,,-89,,889,,897||Asia,,Asia,,29.84,,89.29,,500,,504||Europe,,Europe,,51,,17.5,,506,,512||Africa,,Africa,,7.18,,21.09,,517,,523|| 38 | United States,,United States,,40,,-100,,35,,48|| 39 | Mexico,,Mexico,,23,,-102,,14,,20||United States,,United States,,40,,-100,,381,,394||New Zealand,,New Zealand,,-42,,174,,993,,1004|| 40 | 41 | Israel,,Israel,,31,,35,,278,,284||Gaza,,Gaza,,31.41,,34.33,,303,,307||Egypt,,Egypt,,26,,30,,313,,318||Gaza,,Gaza,,31.41,,34.33,,462,,466|| 42 | Kenya,,Kenya,,1,,38,,79,,84||Kenya,,Kenya,,1,,38,,145,,150||Kisumu,,Kisumu,,-0.1,,34.75,,266,,272||India,,India,,21,,78,,376,,381||Britain,,Britain,,55,,-3,,810,,817||London,,London,,51.50,,-0.12,,872,,878||Kenya,,Kenya,,1,,38,,1436,,1441|| 43 | United States,,United States,,40,,-100,,30,,43||Norway,,Norway,,61,,8,,48,,54||Norway,,Norway,,61,,8,,298,,304||North Carolina,,North Carolina,,35.5,,-80,,578,,592||Mexico,,Mexico,,23,,-102,,1404,,1410||China,,China,,35,,103,,1412,,1417||Japan,,Japan,,35,,136,,1419,,1424||Brazil,,Brazil,,-10,,-52,,1438,,1444|| 44 | India,,India,,21,,78,,327,,332||Sudan,,Sudan,,15,,32,,337,,342||Morocco,,Morocco,,32,,-6,,365,,372||Nigeria,,Nigeria,,8,,10,,392,,399||Egypt,,Egypt,,26,,30,,1214,,1219||Cairo,,Cairo,,30.04,,31.23,,1261,,1266||Africa,,Africa,,7.18,,21.09,,1397,,1403||Saudi Arabia,,Saudi Arabia,,24,,45,,1452,,1464||Iran,,Iran,,32,,53,,1479,,1483|| 45 | 46 | United States,,United States,,40,,-100,,1158,,1171|| 47 | New York,,New York,,43,,-75,,313,,321|| 48 | 49 | 50 | Argentina,,Argentina,,-34,,-64,,36,,45||Buenos Aires,,Buenos Aires,,-34.60,,-58.38,,254,,266||Argentina,,Argentina,,-34,,-64,,386,,395||Argentina,,Argentina,,-34,,-64,,655,,664||South America,,South America,,-14.60,,-57.65,,716,,729||Argentina,,Argentina,,-34,,-64,,893,,902|| 51 | United Kingdom,,United Kingdom,,55,,-3,,298,,312||Kenya,,Kenya,,1,,38,,1100,,1105||Kenya,,Kenya,,1,,38,,1196,,1201|| 52 | Africa,,Africa,,7.18,,21.09,,105,,111||Zimbabwe,,Zimbabwe,,-20,,30,,203,,211||Angola,,Angola,,-12.5,,18.5,,873,,879|| 53 | Ethiopia,,Ethiopia,,8,,38,,202,,210||Sudan,,Sudan,,15,,32,,215,,220||Africa,,Africa,,7.18,,21.09,,411,,417|| 54 | Egypt,,Egypt,,26,,30,,110,,115||United States,,United States,,40,,-100,,803,,816|| 55 | Asia,,Asia,,29.84,,89.29,,88,,92||Afghanistan,,Afghanistan,,33,,65,,459,,470||India,,India,,21,,78,,472,,477||Pakistan,,Pakistan,,30,,70,,483,,491||Asia,,Asia,,29.84,,89.29,,624,,628|| 56 | Peru,,Peru,,-10,,-76,,25,,29||Peru,,Peru,,-10,,-76,,516,,520|| 57 | South Africa,,South Africa,,-30,,25,,28,,40||Johannesburg,,Johannesburg,,-26.20,,28.04,,763,,775||Mozambique,,Mozambique,,-18.25,,35,,940,,950||United States,,United States,,40,,-100,,1344,,1357|| 58 | Dallas County,,Dallas County,,32.77,,-96.78,,409,,422|| 59 | Libya,,Libya,,27,,17,,1,,6||Bulgaria,,Bulgaria,,42.75,,25.5,,265,,273||Libya,,Libya,,27,,17,,648,,653||Libya,,Libya,,27,,17,,800,,805||Libya,,Libya,,27,,17,,912,,917||Libya,,Libya,,27,,17,,993,,998||Libya,,Libya,,27,,17,,1276,,1281||Libya,,Libya,,27,,17,,1369,,1374||Libya,,Libya,,27,,17,,1395,,1400||France,,France,,47,,2,,1558,,1564||France,,France,,47,,2,,1614,,1620||Libya,,Libya,,27,,17,,1635,,1640||Libya,,Libya,,27,,17,,1670,,1675||Europe,,Europe,,51,,17.5,,1770,,1776||Paris,,Paris,,48.85,,2.35,,1875,,1880||Bulgaria,,Bulgaria,,42.75,,25.5,,2175,,2183|| 60 | France,,France,,47,,2,,878,,884|| 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 | 85 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | 127 | 128 | 129 | 130 | 131 | 132 | 133 | 134 | 135 | 136 | 137 | 138 | United States,,United States,,40,,-100,,196,,209||Canada,,Canada,,60,,-95,,214,,220||Honduras,,Honduras,,15,,-86.5,,341,,349||United States,,United States,,40,,-100,,504,,517||United States,,United States,,40,,-100,,583,,596||Canada,,Canada,,60,,-95,,624,,630||United States,,United States,,40,,-100,,750,,763||Arizona,,Arizona,,34,,-112,,801,,808||California,,California,,37,,-120,,810,,820||Colorado,,Colorado,,39,,-105.5,,822,,830||Georgia,,Georgia,,33,,-83.5,,832,,839||Illinois,,Illinois,,40,,-89,,841,,849||Missouri,,Missouri,,38.5,,-92.5,,851,,859||New Jersey,,New Jersey,,40,,-74.5,,861,,871||New Mexico,,New Mexico,,34,,-106,,873,,883||New York,,New York,,43,,-75,,885,,893||Ohio,,Ohio,,40.5,,-82.5,,895,,899||Oklahoma,,Oklahoma,,35.5,,-98,,901,,909||Oregon,,Oregon,,44,,-120.5,,911,,917||Tennessee,,Tennessee,,36,,-86,,919,,928||Utah,,Utah,,39,,-111,,930,,934||Washington,,Washington,,47.5,,-120.5,,936,,946||Wisconsin,,Wisconsin,,44.5,,-89.5,,951,,960||Canada,,Canada,,60,,-95,,975,,981||Alberta,,Alberta,,55,,-115,,1012,,1019||Manitoba,,Manitoba,,55,,-97,,1021,,1029||Ontario,,Ontario,,50,,-85,,1031,,1038||Honduras,,Honduras,,15,,-86.5,,1629,,1637||United States,,United States,,40,,-100,,1651,,1664||Honduras,,Honduras,,15,,-86.5,,1722,,1730||Honduras,,Honduras,,15,,-86.5,,1910,,1918|| 139 | United States,,United States,,40,,-100,,112,,125||Canada,,Canada,,60,,-95,,130,,136||Canada,,Canada,,60,,-95,,280,,286|| 140 | United States,,United States,,40,,-100,,181,,194||Georgia,,Georgia,,33,,-83.5,,903,,910||United States,,United States,,40,,-100,,933,,946|| 141 | Belgium,,Belgium,,50.83,,4,,225,,232||Belgium,,Belgium,,50.83,,4,,837,,844||Belgium,,Belgium,,50.83,,4,,1083,,1090||Belgium,,Belgium,,50.83,,4,,1199,,1206||New Zealand,,New Zealand,,-42,,174,,1601,,1612|| 142 | South Africa,,South Africa,,-30,,25,,79,,91||Johannesburg,,Johannesburg,,-26.20,,28.04,,946,,958|| 143 | 144 | Wisconsin,,Wisconsin,,44.5,,-89.5,,119,,128||Michigan,,Michigan,,44,,-85,,229,,237||Oregon,,Oregon,,44,,-120.5,,239,,245||New Mexico,,New Mexico,,34,,-106,,247,,257||New York,,New York,,43,,-75,,259,,267||Indiana,,Indiana,,40,,-86,,269,,276||Ohio,,Ohio,,40.5,,-82.5,,278,,282||Wisconsin,,Wisconsin,,44.5,,-89.5,,284,,293||Idaho,,Idaho,,45,,-114,,295,,300||Connecticut,,Connecticut,,41.6,,-72.7,,302,,313||Kentucky,,Kentucky,,37.5,,-85,,315,,323||Washington,,Washington,,47.5,,-120.5,,534,,544||Pennsylvania,,Pennsylvania,,41,,-77.5,,546,,558||California,,California,,37,,-120,,563,,573||Wisconsin,,Wisconsin,,44.5,,-89.5,,1903,,1912||Wisconsin,,Wisconsin,,44.5,,-89.5,,1933,,1942||California,,California,,37,,-120,,2633,,2643|| 145 | United States,,United States,,40,,-100,,85,,98||Canada,,Canada,,60,,-95,,103,,109||California,,California,,37,,-120,,310,,320|| 146 | Pennsylvania,,Pennsylvania,,41,,-77.5,,381,,393||New Jersey,,New Jersey,,40,,-74.5,,399,,409||Canada,,Canada,,60,,-95,,513,,519||Canada,,Canada,,60,,-95,,655,,661||Canada,,Canada,,60,,-95,,736,,742||Canada,,Canada,,60,,-95,,928,,934||Pennsylvania,,Pennsylvania,,41,,-77.5,,1316,,1328||New York,,New York,,43,,-75,,1516,,1524||Canada,,Canada,,60,,-95,,1658,,1664||Texas,,Texas,,31,,-100,,1821,,1826|| 147 | Dominican Republic,,Dominican Republic,,19,,-70.66,,219,,237||Jamaica,,Jamaica,,18,,-77,,242,,249||Dominican Republic,,Dominican Republic,,19,,-70.66,,1044,,1062||Jamaica,,Jamaica,,18,,-77,,1066,,1073||Dominican Republic,,Dominican Republic,,19,,-70.66,,1666,,1684||Haiti,,Haiti,,19,,-72.41,,1888,,1893|| 148 | 149 | 150 | Uganda,,Uganda,,1,,32,,798,,804|| 151 | Ontario,,Ontario,,50,,-85,,680,,687||Ontario,,Ontario,,50,,-85,,767,,774||New York,,New York,,43,,-75,,1360,,1368||United States,,United States,,40,,-100,,1376,,1389||Ontario,,Ontario,,50,,-85,,1439,,1446|| 152 | 153 | North America,,North America,,54.77,,-105.64,,3144,,3157||United States,,United States,,40,,-100,,3200,,3213||United States,,United States,,40,,-100,,3543,,3556||Italy,,Italy,,43,,12,,5203,,5208||Rome,,Rome,,41.9,,12.5,,5691,,5695||Singapore,,Singapore,,1.3,,103.8,,7140,,7149||Santa Maria,,Santa Maria,,6.55,,125.46,,7275,,7286||Pung-Pang,,Pongpong,,6.49,,125.46,,7527,,7535|| 154 | India,,India,,21,,78,,186,,191|| 155 | Singapore,,Singapore,,1.3,,103.8,,152,,161||Hong Kong,,Hong Kong,,22.3,,114.2,,466,,475|| 156 | 157 | New South Wales,,New South Wales,,-32.16,,147.01,,87,,102||Sydney,,Sydney,,-33.86,,151.20,,772,,778||Sydney,,Sydney,,-33.86,,151.20,,1067,,1073||Melbourne,,Melbourne,,-37.81,,144.96,,1078,,1087|| 158 | Africa,,Africa,,7.18,,21.09,,336,,342||France,,France,,47,,2,,409,,415||Germany,,Germany,,51,,9,,420,,427||France,,France,,47,,2,,567,,573||Germany,,Germany,,51,,9,,578,,585|| 159 | Angola,,Angola,,-12.5,,18.5,,83,,89||London,,London,,51.50,,-0.12,,793,,799||Munich,,Munich,,48.13,,11.56,,804,,810||Angola,,Angola,,-12.5,,18.5,,1314,,1320|| 160 | Zimbabwe,,Zimbabwe,,-20,,30,,46,,54||Botswana,,Botswana,,-24.65,,25.90,,1186,,1194||Mozambique,,Mozambique,,-18.25,,35,,1196,,1206||Zimbabwe,,Zimbabwe,,-20,,30,,1237,,1245||South Africa,,South Africa,,-30,,25,,1212,,1224||Zimbabwe,,Zimbabwe,,-20,,30,,1519,,1527||Zimbabwe,,Zimbabwe,,-20,,30,,1602,,1610|| 161 | 162 | London,,London,,51.50,,-0.12,,929,,935||Italy,,Italy,,43,,12,,1053,,1058||France,,France,,47,,2,,1218,,1224|| 163 | United States,,United States,,40,,-100,,41,,54||Canada,,Canada,,60,,-95,,82,,88||Alberta,,Alberta,,55,,-115,,895,,902||Manitoba,,Manitoba,,55,,-97,,918,,926||Calgary,,Calgary,,51.05,,-114.06,,1381,,1388|| 164 | India,,India,,21,,78,,591,,596||Malaysia,,Malaysia,,2.5,,112.5,,645,,653||Myanmar,,Myanmar,,22,,96,,655,,662||Bangladesh,,Bangladesh,,23.8,,90.3,,664,,674||Tanzania,,Tanzania,,-6.30,,34.85,,699,,707||Kenya,,Kenya,,1,,38,,712,,717|| 165 | Angola,,Angola,,-12.5,,18.5,,37,,43||Congo,,Zaire,,-2.88,,23.65,,714,,719||Uige,,Uige,,-7.61,,15.05,,1055,,1059||Uige,,Uige,,-7.61,,15.05,,1677,,1681||Angola,,Angola,,-12.5,,18.5,,1748,,1754||Mozambique,,Mozambique,,-18.25,,35,,1794,,1804||Angola,,Angola,,-12.5,,18.5,,2236,,2242||Angola,,Angola,,-12.5,,18.5,,2455,,2461|| 166 | Indonesia,,Indonesia,,-5,,120,,1353,,1362|| 167 | Zimbabwe,,Zimbabwe,,-20,,30,,96,,104||Africa,,Africa,,7.18,,21.09,,188,,194||Zimbabwe,,Zimbabwe,,-20,,30,,529,,537||Mozambique,,Mozambique,,-18.25,,35,,769,,779||South Africa,,South Africa,,-30,,25,,781,,793|| 168 | Sydney,,Sydney,,-33.86,,151.20,,85,,91||Melbourne,,Melbourne,,-37.81,,144.96,,1546,,1555|| 169 | Canada,,Canada,,60,,-95,,1022,,1028||Nicaragua,,Nicaragua,,13.09,,-86.00,,1598,,1607||Haiti,,Haiti,,19,,-72.41,,1657,,1662||Dominican Republic,,Dominican Republic,,19,,-70.66,,1671,,1689|| 170 | Chad,,Chad,,15,,19,,511,,515||Guinea,,Guinea,,11,,-10,,517,,523||Mali,,Mali,,17,,-4,,534,,538||Mauritania,,Mauritania,,20,,-12,,540,,550||Senegal,,Senegal,,14,,-14,,552,,559||Benin,,Benin,,6.46,,2.6,,640,,645||Cape Verde,,Cape Verde,,15.11,,-23.61,,647,,657||Ghana,,Ghana,,7.81,,-1.05,,708,,713||Niger,,Niger,,16,,8,,730,,735||Nigeria,,Nigeria,,8,,10,,737,,744|| 171 | United States,,United States,,40,,-100,,877,,890||Canada,,Canada,,60,,-95,,1309,,1315||Georgia,,Georgia,,33,,-83.5,,1464,,1471||Oklahoma,,Oklahoma,,35.5,,-98,,1473,,1481||Pennsylvania,,Pennsylvania,,41,,-77.5,,1483,,1495||Wisconsin,,Wisconsin,,44.5,,-89.5,,1497,,1506||California,,California,,37,,-120,,1511,,1521||California,,California,,37,,-120,,1700,,1710|| 172 | 173 | Sweden,,Sweden,,63,,16,,82,,88||Stockholm,,Stockholm,,59.32,,18.06,,140,,149||United Kingdom,,United Kingdom,,55,,-3,,636,,650||Sweden,,Sweden,,63,,16,,753,,759||Sweden,,Sweden,,63,,16,,1094,,1100|| 174 | United Kingdom,,United Kingdom,,55,,-3,,690,,704|| 175 | Germany,,Germany,,51,,9,,90,,97||Belgium,,Belgium,,50.83,,4,,926,,933||France,,France,,47,,2,,935,,941|| 176 | Pakistan,,Pakistan,,30,,70,,231,,239||Pakistan,,Pakistan,,30,,70,,1089,,1097||Australia,,Australia,,-25,,133,,1180,,1189||France,,France,,47,,2,,1191,,1197||Japan,,Japan,,35,,136,,1199,,1204||China,,China,,35,,103,,1214,,1219||Russia,,Russia,,60,,90,,1221,,1227||Iran,,Iran,,32,,53,,1229,,1233||Syria,,Syria,,35,,38,,1235,,1240|| 177 | Mexico,,Mexico,,23,,-102,,174,,180||United States,,United States,,40,,-100,,189,,202||China,,China,,35,,103,,286,,291||Japan,,Japan,,35,,136,,293,,298||United Kingdom,,United Kingdom,,55,,-3,,308,,322|| 178 | Ceres,,Ceres,,37.60,,-120.95,,82,,87||California,,California,,37,,-120,,89,,99||United States,,United States,,40,,-100,,107,,120||United States,,United States,,40,,-100,,722,,735||Mississauga,,Mississauga,,43.6,,-79.65,,1835,,1846||Ontario,,Ontario,,50,,-85,,1848,,1855||Canada,,Canada,,60,,-95,,1859,,1865||China,,China,,35,,103,,2106,,2111|| 179 | Georgia,,Georgia,,33,,-83.5,,41,,48||Angola,,Angola,,-12.5,,18.5,,406,,412||Uganda,,Uganda,,1,,32,,448,,454|| 180 | Kenya,,Kenya,,1,,38,,105,,110||Kenya,,Kenya,,1,,38,,268,,273||Kenya,,Kenya,,1,,38,,473,,478||Indian Ocean,,Indian Ocean,,-20,,80,,554,,566|| 181 | Santa Ana,,Santa Ana,,33.74,,-117.88,,229,,238||Orange County,,Orange County,,33.67,,-117.78,,1640,,1653||Santa Ana,,Santa Ana,,33.74,,-117.88,,4722,,4731||Santa Ana,,Santa Ana,,33.74,,-117.88,,5238,,5247||Santa Ana,,Santa Ana,,33.74,,-117.88,,5581,,5590||Santa Ana,,Santa Ana,,33.74,,-117.88,,6149,,6158||Santa Ana,,Santa Ana,,33.74,,-117.88,,6377,,6386||Newport Beach,,Newport Beach,,33.61,,-117.89,,7056,,7069|| 182 | Papua New Guinea,,Papua New Guinea,,-6,,147,,200,,216||Malaysia,,Malaysia,,2.5,,112.5,,432,,440||Vietnam,,Vietnam,,16.16,,107.83,,442,,449|| 183 | Mayotte,,Mayotte,,-12.84,,45.13,,636,,643|| 184 | Africa,,Africa,,7.18,,21.09,,146,,152||United States,,United States,,40,,-100,,868,,881||France,,France,,47,,2,,883,,889||Belgium,,Belgium,,50.83,,4,,891,,898||Australia,,Australia,,-25,,133,,900,,909||Denmark,,Denmark,,56,,10,,948,,955||Africa,,Africa,,7.18,,21.09,,2103,,2109||Africa,,Africa,,7.18,,21.09,,2142,,2148|| 185 | California,,California,,37,,-120,,434,,444||California,,California,,37,,-120,,678,,688||California,,California,,37,,-120,,904,,914|| 186 | Zimbabwe,,Zimbabwe,,-20,,30,,1242,,1250|| 187 | Midrand,,Midrand,,-25.99,,28.12,,1307,,1314||Johannesburg,,Johannesburg,,-26.20,,28.04,,1346,,1358||China,,China,,35,,103,,2390,,2395||China,,China,,35,,103,,2777,,2782||China,,China,,35,,103,,2792,,2797||South Africa,,South Africa,,-30,,25,,3030,,3042|| 188 | United States,,United States,,40,,-100,,270,,283||Egypt,,Egypt,,26,,30,,354,,359||Bulgaria,,Bulgaria,,42.75,,25.5,,361,,369||Nicaragua,,Nicaragua,,13.09,,-86.00,,375,,384||Lebanon,,Lebanon,,33.83,,35.83,,423,,430||Australia,,Australia,,-25,,133,,1218,,1227||Australia,,Australia,,-25,,133,,1592,,1601||Singapore,,Singapore,,1.3,,103.8,,2279,,2288||Singapore,,Singapore,,1.3,,103.8,,2455,,2464||Rio de Janeiro,,Rio de Janeiro,,-22.90,,-43.19,,2941,,2955||Santa Catarina,,Santa Catarina,,-27.25,,-50.33,,2965,,2979||Canada,,Canada,,60,,-95,,3536,,3542||Chile,,Chile,,-30,,-71,,3760,,3765||Chile,,Chile,,-30,,-71,,4070,,4075||New York,,New York,,40.71,,-74.00,,5100,,5108||Japan,,Japan,,35,,136,,5506,,5511||Spain,,Spain,,40,,-4,,5516,,5521||United States,,United States,,40,,-100,,5904,,5917||Iraq,,Iraq,,33,,44,,5921,,5925||Turkey,,Turkey,,39,,35,,5958,,5964||New York,,New York,,40.71,,-74.00,,6069,,6077||Toronto,,Toronto,,43.7,,-79.4,,6082,,6089||Istanbul,,Istanbul,,41.01,,28.95,,6345,,6353||United Kingdom,,United Kingdom,,55,,-3,,6425,,6439||England,,England,,51.5,,-0.11,,6457,,6464||Scotland,,Scotland,,56,,-4,,6805,,6813||United States,,United States,,40,,-100,,7443,,7456||Wisconsin,,Wisconsin,,44.5,,-89.5,,7726,,7735||United States,,United States,,40,,-100,,8046,,8059|| 189 | Panama,,Panama,,9,,-80,,259,,265||Australia,,Australia,,-25,,133,,1122,,1131||Asia,,Asia,,29.84,,89.29,,1136,,1140|| 190 | Australia,,Australia,,-25,,133,,119,,128|| 191 | Cellardyke,,Cellardyke,,56.21,,-2.7,,83,,93||Scotland,,Scotland,,56,,-4,,103,,111||Scotland,,Scotland,,56,,-4,,222,,230||Scotland,,Scotland,,56,,-4,,3119,,3127|| 192 | Mexico,,Mexico,,23,,-102,,758,,764||United States,,United States,,40,,-100,,773,,786||Canada,,Canada,,60,,-95,,1004,,1010||Nova Scotia,,Nova Scotia,,45,,-63,,1079,,1090||Ontario,,Ontario,,50,,-85,,1102,,1109||Alberta,,Alberta,,55,,-115,,1120,,1127||New Brunswick,,New Brunswick,,46,,-66,,1136,,1149||Quebec,,Quebec,,53,,-70,,1162,,1168||Alberta,,Alberta,,55,,-115,,1192,,1199||Mexico,,Mexico,,23,,-102,,1324,,1330||Mexico,,Mexico,,23,,-102,,1864,,1870||Egypt,,Egypt,,26,,30,,1910,,1915||Mexico,,Mexico,,23,,-102,,2123,,2129||Germany,,Germany,,51,,9,,2162,,2169||Mexico,,Mexico,,23,,-102,,2250,,2256||Spain,,Spain,,40,,-4,,2299,,2304||Mexico,,Mexico,,23,,-102,,2405,,2411||Mexico,,Mexico,,23,,-102,,2528,,2534||Hong Kong,,Hong Kong,,22.3,,114.2,,2413,,2422||Hong Kong,,Hong Kong,,22.3,,114.2,,3011,,3020||China,,China,,35,,103,,3263,,3268||Mexico,,Mexico,,23,,-102,,3292,,3298||Mexico,,Mexico,,23,,-102,,3721,,3727||New Zealand,,New Zealand,,-42,,174,,5241,,5252||North America,,North America,,54.77,,-105.64,,5533,,5546||Nigeria,,Nigeria,,8,,10,,6044,,6051||Nigeria,,Nigeria,,8,,10,,6138,,6145||Delaware,,Delaware,,39,,-75.5,,6547,,6555||Illinois,,Illinois,,40,,-89,,6557,,6565||Tennessee,,Tennessee,,36,,-86,,6597,,6606||Texas,,Texas,,31,,-100,,6859,,6864||Alabama,,Alabama,,32.7,,-86.7,,6885,,6892||Mexico,,Mexico,,23,,-102,,7214,,7220||Mexico,,Mexico,,23,,-102,,7366,,7372|| 193 | Pakistan,,Pakistan,,30,,70,,38,,46||Pakistan,,Pakistan,,30,,70,,675,,683||Pakistan,,Pakistan,,30,,70,,1014,,1022||Pakistan,,Pakistan,,30,,70,,1873,,1881||Malaysia,,Malaysia,,2.5,,112.5,,2926,,2934||Pakistan,,Pakistan,,30,,70,,3012,,3020||Britain,,Britain,,55,,-3,,3026,,3033|| 194 | Rwanda,,Rwanda,,-1.94,,29.87,,3241,,3247|| 195 | Kenya,,Kenya,,1,,38,,919,,924||Kenya,,Kenya,,1,,38,,1026,,1031||Kenya,,Kenya,,1,,38,,1487,,1492||Uganda,,Uganda,,1,,32,,1770,,1776|| 196 | 197 | United States,,United States,,40,,-100,,27,,40||United States,,United States,,40,,-100,,294,,307||Iraq,,Iraq,,33,,44,,439,,443||Afghanistan,,Afghanistan,,33,,65,,445,,456||North Korea,,North Korea,,40,,127,,458,,469||Iran,,Iran,,32,,53,,559,,563||Mexico,,Mexico,,23,,-102,,923,,929|| 198 | Indonesia,,Indonesia,,-5,,120,,1392,,1401|| 199 | Mexico,,Mexico,,23,,-102,,48,,54||Australia,,Australia,,-25,,133,,253,,262||Brazil,,Brazil,,-10,,-52,,264,,270||France,,France,,47,,2,,272,,278||Italy,,Italy,,43,,12,,280,,285||New Zealand,,New Zealand,,-42,,174,,287,,298||Norway,,Norway,,61,,8,,300,,306||Switzerland,,Switzerland,,46.83,,8.33,,308,,319||Brazil,,Brazil,,-10,,-52,,799,,805||Brazil,,Brazil,,-10,,-52,,1092,,1098||United States,,United States,,40,,-100,,1117,,1130||Argentina,,Argentina,,-34,,-64,,1218,,1227||China,,China,,35,,103,,1346,,1351||China,,China,,35,,103,,1913,,1918||China,,China,,35,,103,,2208,,2213||India,,India,,21,,78,,2723,,2728||Kenya,,Kenya,,1,,38,,3616,,3621||Mexico,,Mexico,,23,,-102,,4500,,4506||United Kingdom,,United Kingdom,,55,,-3,,5041,,5055||Britain,,Britain,,55,,-3,,6139,,6146||United States,,United States,,40,,-100,,6836,,6849||Vietnam,,Vietnam,,16.16,,107.83,,7067,,7074|| 200 | Uganda,,Uganda,,1,,32,,136,,142||Uganda,,Uganda,,1,,32,,471,,477||Uganda,,Uganda,,1,,32,,542,,548||Uganda,,Uganda,,1,,32,,591,,597|| 201 | 202 | China,,China,,35,,103,,398,,403|| 203 | Northern Ireland,,Northern Ireland,,54.5,,-6.5,,38,,54||Northern Ireland,,Northern Ireland,,54.5,,-6.5,,243,,259||Spain,,Spain,,40,,-4,,353,,358||England,,England,,51.5,,-0.11,,385,,392||Northern Ireland,,Northern Ireland,,54.5,,-6.5,,461,,477|| 204 | Denmark,,Denmark,,56,,10,,257,,264|| 205 | Las Vegas,,Las Vegas,,36.17,,-115.13,,309,,318|| 206 | 207 | Acapulco,,Acapulco,,16.86,,-99.88,,153,,161||Mexico,,Mexico,,23,,-102,,1597,,1603||Mexico,,Mexico,,23,,-102,,2109,,2115||Mexico City,,Mexico City,,19.43,,-99.13,,2151,,2162||South America,,South America,,-14.60,,-57.65,,2222,,2235||United States,,United States,,40,,-100,,2244,,2257||Mexico,,Mexico,,23,,-102,,2412,,2418|| 208 | New Orleans,,New Orleans,,29.95,,-90.06,,22,,33||New Orleans,,New Orleans,,29.95,,-90.06,,1128,,1139|| 209 | 210 | India,,India,,21,,78,,105,,110|| 211 | Australia,,Australia,,-25,,133,,60,,69||Australia,,Australia,,-25,,133,,161,,170||Australia,,Australia,,-25,,133,,521,,530||Australia,,Australia,,-25,,133,,643,,652||Australia,,Australia,,-25,,133,,1066,,1075||Indonesia,,Indonesia,,-5,,120,,2296,,2305|| 212 | Angola,,Angola,,-12.5,,18.5,,311,,317||Angola,,Angola,,-12.5,,18.5,,376,,382||Angola,,Angola,,-12.5,,18.5,,2079,,2085||Angola,,Angola,,-12.5,,18.5,,2157,,2163|| 213 | China,,China,,35,,103,,67,,72||United Kingdom,,United Kingdom,,55,,-3,,673,,687||China,,China,,35,,103,,1729,,1734||China,,China,,35,,103,,1792,,1797||United States,,United States,,40,,-100,,2186,,2199|| 214 | Iraq,,Iraq,,33,,44,,141,,145||Iraq,,Iraq,,33,,44,,454,,458||Kuwait,,Kuwait,,29.5,,45.75,,468,,474|| 215 | 216 | Birmingham,,Birmingham,,33.52,,-86.81,,891,,901|| 217 | 218 | Bradford,,Bradford,,44.11,,-79.56,,870,,878||Ontario,,Ontario,,50,,-85,,880,,887|| 219 | Peru,,Peru,,-10,,-76,,132,,136||Bolivia,,Bolivia,,-16.71,,-64.66,,154,,161||Massachusetts,,Massachusetts,,42.3,,-71.8,,1475,,1488||London,,London,,51.50,,-0.12,,2065,,2071|| 220 | 221 | Australia,,Australia,,-25,,133,,37,,46|| 222 | Britain,,Britain,,55,,-3,,85,,92||Chechnya,,Chechnya,,43.4,,45.71,,632,,640||Moscow,,Moscow,,55.75,,37.61,,662,,668||Muswell Hill,,Muswell Hill,,51.59,,-0.14,,1949,,1961||Moscow,,Moscow,,55.75,,37.61,,2479,,2485|| 223 | China,,China,,35,,103,,113,,118|| 224 | United Kingdom,,United Kingdom,,55,,-3,,415,,429||Baghdad,,Baghdad,,33.33,,44.38,,1614,,1621|| 225 | Georgia,,Georgia,,42,,43.5,,1,,8||Georgia,,Georgia,,42,,43.5,,464,,471||Georgia,,Georgia,,42,,43.5,,1432,,1439|| 226 | Las Vegas,,Las Vegas,,36.17,,-115.13,,936,,945|| 227 | Camborne,,Camborne,,50.21,,-5.3,,275,,283|| 228 | Denver,,Denver,,39.76,,-104.88,,166,,172||Peru,,Peru,,-10,,-76,,249,,253||San Juan,,San Juan,,-7.29,,-78.49,,442,,450||Peru,,Peru,,-10,,-76,,1408,,1412||Indonesia,,Indonesia,,-5,,120,,2644,,2653||United States,,United States,,40,,-100,,2758,,2771|| 229 | United States,,United States,,40,,-100,,78,,91||Florida,,Florida,,28.1,,-81.6,,1327,,1334||Utah,,Utah,,39,,-111,,1387,,1391||United States,,United States,,40,,-100,,1711,,1724||United States,,United States,,40,,-100,,2574,,2587|| 230 | -------------------------------------------------------------------------------- /data/iaa_answers.txt: -------------------------------------------------------------------------------- 1 | https://en.wikipedia.org/wiki/Dorset 2 | 96 3 | Dorset 4 | https://en.wikipedia.org/wiki/England 5 | 104 6 | England 7 | https://en.wikipedia.org/wiki/United_Kingdom 8 | 390 9 | United Kingdom 10 | https://en.wikipedia.org/wiki/Michigan 11 | 98 12 | Michigan 13 | https://en.wikipedia.org/wiki/Lake_Erie 14 | 123 15 | Lake Erie 16 | https://en.wikipedia.org/wiki/Pointe_Mouillee_State_Game_Area 17 | 142 18 | Mouillee 19 | https://en.wikipedia.org/wiki/Monroe_County,_Michigan 20 | 170 21 | Monroe County 22 | https://en.wikipedia.org/wiki/Michigan 23 | 262 24 | Michigan 25 | https://en.wikipedia.org/wiki/Ames,_Iowa 26 | 397 27 | Ames 28 | https://en.wikipedia.org/wiki/Iowa 29 | 403 30 | Iowa 31 | https://en.wikipedia.org/wiki/North_America 32 | 574 33 | North America 34 | https://en.wikipedia.org/wiki/Asia 35 | 675 36 | Asia 37 | https://en.wikipedia.org/wiki/Michigan 38 | 921 39 | Michigan 40 | https://en.wikipedia.org/wiki/United_States 41 | 956 42 | United States 43 | https://en.wikipedia.org/wiki/Buckinghamshire 44 | 85 45 | Buckinghamshire 46 | https://en.wikipedia.org/wiki/England 47 | 102 48 | England 49 | https://en.wikipedia.org/wiki/Milton_Keynes 50 | 333 51 | Milton Keynes 52 | https://en.wikipedia.org/wiki/Peru 53 | 24 54 | Peru 55 | https://en.wikipedia.org/wiki/Peru 56 | 174 57 | Peru 58 | https://en.wikipedia.org/wiki/Peru 59 | 338 60 | Peru 61 | https://en.wikipedia.org/wiki/Peru 62 | 515 63 | Peru 64 | https://en.wikipedia.org/wiki/Libya 65 | 0 66 | Libya 67 | https://en.wikipedia.org/wiki/Sofia 68 | 223 69 | Sofia 70 | https://en.wikipedia.org/wiki/Bulgaria 71 | 230 72 | Bulgaria 73 | https://en.wikipedia.org/wiki/Bulgaria 74 | 264 75 | Bulgaria 76 | https://en.wikipedia.org/wiki/Benghazi 77 | 547 78 | Benghazi 79 | https://en.wikipedia.org/wiki/Libya 80 | 647 81 | Libya 82 | https://en.wikipedia.org/wiki/Libya 83 | 799 84 | Libya 85 | https://en.wikipedia.org/wiki/Libya 86 | 911 87 | Libya 88 | https://en.wikipedia.org/wiki/Qatar 89 | 930 90 | Qatar 91 | https://en.wikipedia.org/wiki/Libya 92 | 992 93 | Libya 94 | https://en.wikipedia.org/wiki/Benghazi 95 | 1135 96 | Benghazi 97 | https://en.wikipedia.org/wiki/Libya 98 | 1275 99 | Libya 100 | https://en.wikipedia.org/wiki/Libya 101 | 1368 102 | Libya 103 | https://en.wikipedia.org/wiki/Libya 104 | 1394 105 | Libya 106 | https://en.wikipedia.org/wiki/France 107 | 1557 108 | France 109 | https://en.wikipedia.org/wiki/France 110 | 1613 111 | France 112 | https://en.wikipedia.org/wiki/Libya 113 | 1634 114 | Libya 115 | https://en.wikipedia.org/wiki/Libya 116 | 1669 117 | Libya 118 | https://en.wikipedia.org/wiki/France 119 | 1780 120 | France 121 | https://en.wikipedia.org/wiki/Libya 122 | 1837 123 | Libya 124 | https://en.wikipedia.org/wiki/Paris 125 | 1874 126 | Paris 127 | https://en.wikipedia.org/wiki/Bulgaria 128 | 2174 129 | Bulgaria 130 | https://en.wikipedia.org/wiki/London 131 | 80 132 | London 133 | https://en.wikipedia.org/wiki/United_Kingdom 134 | 612 135 | UK 136 | https://en.wikipedia.org/wiki/United_Kingdom 137 | 738 138 | UK 139 | https://en.wikipedia.org/wiki/United_States 140 | 163 141 | U.S. 142 | https://en.wikipedia.org/wiki/Central_Plateau_(Haiti) 143 | 710 144 | Central Plateau 145 | https://en.wikipedia.org/wiki/Artibonite_(department) 146 | 730 147 | Artibonite 148 | https://en.wikipedia.org/wiki/Port-au-Prince 149 | 803 150 | Port-Au-Prince 151 | https://en.wikipedia.org/wiki/Java 152 | 80 153 | Java 154 | https://en.wikipedia.org/wiki/Java 155 | 340 156 | Java 157 | https://en.wikipedia.org/wiki/Pangandaran 158 | 935 159 | Pangandaran 160 | https://en.wikipedia.org/wiki/Indonesia 161 | 1226 162 | Indonesia 163 | https://en.wikipedia.org/wiki/Hong_Kong 164 | 54 165 | Hong Kong 166 | https://en.wikipedia.org/wiki/Kitaakita 167 | 89 168 | Kitaakita 169 | https://en.wikipedia.org/wiki/Akita_Prefecture 170 | 114 171 | Akita Prefecture 172 | https://en.wikipedia.org/wiki/Japan 173 | 143 174 | Japan 175 | https://en.wikipedia.org/wiki/United_States 176 | 224 177 | United States 178 | https://en.wikipedia.org/wiki/New_York_City 179 | 1201 180 | New York City 181 | https://en.wikipedia.org/wiki/New_York_City 182 | 1313 183 | New York 184 | https://en.wikipedia.org/wiki/California 185 | 1571 186 | California 187 | https://en.wikipedia.org/wiki/Kansas 188 | 1633 189 | Kansas 190 | https://en.wikipedia.org/wiki/Texas 191 | 1644 192 | Texas 193 | https://en.wikipedia.org/wiki/United_States 194 | 2089 195 | U.S. 196 | https://en.wikipedia.org/wiki/Kabanjahe 197 | 71 198 | Kabanjahe 199 | https://en.wikipedia.org/wiki/Sumatra 200 | 93 201 | Sumatra 202 | https://en.wikipedia.org/wiki/Indonesia 203 | 104 204 | Indonesia 205 | https://en.wikipedia.org/wiki/Kabanjahe 206 | 307 207 | Kabanjahe 208 | https://en.wikipedia.org/wiki/Indonesia 209 | 350 210 | Indonesia 211 | https://en.wikipedia.org/wiki/Atlanta 212 | 500 213 | Atlanta 214 | https://en.wikipedia.org/wiki/Georgia_(U.S._state) 215 | 509 216 | Georgia 217 | https://en.wikipedia.org/wiki/Atlanta 218 | 669 219 | Atlanta 220 | https://en.wikipedia.org/wiki/China 221 | 84 222 | China 223 | https://en.wikipedia.org/wiki/Beijing 224 | 120 225 | Beijing 226 | https://en.wikipedia.org/wiki/China 227 | 400 228 | China 229 | https://en.wikipedia.org/wiki/Hebei 230 | 657 231 | Hebei 232 | https://en.wikipedia.org/wiki/Hubei 233 | 786 234 | Hubei 235 | https://en.wikipedia.org/wiki/China 236 | 870 237 | People's Republic of China 238 | https://en.wikipedia.org/wiki/China 239 | 990 240 | China 241 | https://en.wikipedia.org/wiki/Anhui 242 | 1143 243 | Anhui 244 | https://en.wikipedia.org/wiki/Fuyang 245 | 1276 246 | Fuyang 247 | https://en.wikipedia.org/wiki/Anhui 248 | 1286 249 | Anhui 250 | https://en.wikipedia.org/wiki/Beijing 251 | 1531 252 | Beijing 253 | https://en.wikipedia.org/wiki/Anhui 254 | 1581 255 | Anhui 256 | https://en.wikipedia.org/wiki/Guangdong 257 | 1588 258 | Guangdong 259 | https://en.wikipedia.org/wiki/Guangxi 260 | 1599 261 | Guangxi 262 | https://en.wikipedia.org/wiki/Hainan 263 | 1608 264 | Hainan 265 | https://en.wikipedia.org/wiki/Hunan 266 | 1616 267 | Hunan 268 | https://en.wikipedia.org/wiki/Zhejiang 269 | 1623 270 | Zhejiang 271 | https://en.wikipedia.org/wiki/Beijing 272 | 1633 273 | Beijing 274 | https://en.wikipedia.org/wiki/Hubei 275 | 1645 276 | Hubei 277 | https://en.wikipedia.org/wiki/Beijing 278 | 2195 279 | Beijing 280 | https://en.wikipedia.org/wiki/China 281 | 2393 282 | China 283 | https://en.wikipedia.org/wiki/Asia 284 | 80 285 | Asia 286 | https://en.wikipedia.org/wiki/India 287 | 112 288 | India 289 | https://en.wikipedia.org/wiki/Nepal 290 | 119 291 | Nepal 292 | https://en.wikipedia.org/wiki/Bangladesh 293 | 130 294 | Bangladesh 295 | https://en.wikipedia.org/wiki/Bihar 296 | 461 297 | Bihar 298 | https://en.wikipedia.org/wiki/Uttar_Pradesh 299 | 471 300 | Uttar Pradesh 301 | https://en.wikipedia.org/wiki/Bihar 302 | 585 303 | Bihar 304 | https://en.wikipedia.org/wiki/India 305 | 794 306 | India 307 | None 308 | 1112 309 | Gondra 310 | https://en.wikipedia.org/wiki/Uttar_Pradesh 311 | 1131 312 | Uttar Pradesh 313 | https://en.wikipedia.org/wiki/Greenlane 314 | 76 315 | Greenlane 316 | https://en.wikipedia.org/wiki/Auckland 317 | 87 318 | Auckland 319 | https://en.wikipedia.org/wiki/New_Zealand 320 | 97 321 | New Zealand 322 | https://en.wikipedia.org/wiki/Auckland 323 | 252 324 | Auckland 325 | https://en.wikipedia.org/wiki/Greenlane 326 | 1569 327 | Greenlane 328 | https://en.wikipedia.org/wiki/Auckland 329 | 1580 330 | Auckland 331 | https://en.wikipedia.org/wiki/Auckland 332 | 1679 333 | Auckland 334 | https://en.wikipedia.org/wiki/Zimbabwe 335 | 45 336 | Zimbabwe 337 | https://en.wikipedia.org/wiki/Zimbabwe 338 | 293 339 | Zimbabwe 340 | https://en.wikipedia.org/wiki/Zimbabwe 341 | 387 342 | Zimbabwe 343 | https://en.wikipedia.org/wiki/Zimbabwe 344 | 711 345 | Zimbabwe 346 | https://en.wikipedia.org/wiki/Harare 347 | 787 348 | Harare 349 | https://en.wikipedia.org/wiki/Mabvuku 350 | 897 351 | Mabvuku 352 | https://en.wikipedia.org/wiki/Harare 353 | 931 354 | Harare 355 | https://en.wikipedia.org/wiki/Botswana 356 | 1185 357 | Botswana 358 | https://en.wikipedia.org/wiki/Mozambique 359 | 1195 360 | Mozambique 361 | https://en.wikipedia.org/wiki/South_Africa 362 | 1211 363 | South Africa 364 | https://en.wikipedia.org/wiki/Zimbabwe 365 | 1236 366 | Zimbabwe 367 | https://en.wikipedia.org/wiki/Zimbabwe 368 | 1518 369 | Zimbabwe 370 | https://en.wikipedia.org/wiki/Zimbabwe 371 | 1601 372 | Zimbabwe 373 | https://en.wikipedia.org/wiki/Zimbabwe 374 | 3 375 | Zimbabwe 376 | https://en.wikipedia.org/wiki/Wedza_District 377 | 54 378 | Wedza 379 | https://en.wikipedia.org/wiki/Mashonaland_East_Province 380 | 70 381 | Mashonaland East 382 | https://en.wikipedia.org/wiki/Midlands_Province 383 | 549 384 | Midlands 385 | https://en.wikipedia.org/wiki/Zimbabwe 386 | 1241 387 | Zimbabwe 388 | https://en.wikipedia.org/wiki/United_States 389 | 26 390 | United States 391 | https://en.wikipedia.org/wiki/United_States 392 | 293 393 | United States 394 | https://en.wikipedia.org/wiki/Iraq 395 | 438 396 | Iraq 397 | https://en.wikipedia.org/wiki/Afghanistan 398 | 444 399 | Afghanistan 400 | https://en.wikipedia.org/wiki/North_Korea 401 | 457 402 | North Korea 403 | https://en.wikipedia.org/wiki/Iran 404 | 558 405 | Iran 406 | https://en.wikipedia.org/wiki/United_States 407 | 915 408 | U.S. 409 | https://en.wikipedia.org/wiki/Mexico 410 | 922 411 | Mexico 412 | https://en.wikipedia.org/wiki/China 413 | 169 414 | China 415 | https://en.wikipedia.org/wiki/Indonesia 416 | 176 417 | Indonesia 418 | https://en.wikipedia.org/wiki/Africa 419 | 208 420 | Africa 421 | https://en.wikipedia.org/wiki/China 422 | 492 423 | China 424 | https://en.wikipedia.org/wiki/Indonesia 425 | 502 426 | Indonesia 427 | https://en.wikipedia.org/wiki/Africa 428 | 516 429 | Africa 430 | https://en.wikipedia.org/wiki/Rome 431 | 719 432 | Rome 433 | https://en.wikipedia.org/wiki/Italy 434 | 725 435 | Italy 436 | https://en.wikipedia.org/wiki/Africa 437 | 1117 438 | Africa 439 | https://en.wikipedia.org/wiki/Indonesia 440 | 1391 441 | Indonesia 442 | https://en.wikipedia.org/wiki/Java 443 | 1715 444 | Java 445 | https://en.wikipedia.org/wiki/China 446 | 1813 447 | China 448 | https://en.wikipedia.org/wiki/China 449 | 1858 450 | China 451 | https://en.wikipedia.org/wiki/Mexico 452 | 47 453 | Mexico 454 | https://en.wikipedia.org/wiki/Australia 455 | 252 456 | Australia 457 | https://en.wikipedia.org/wiki/Brazil 458 | 263 459 | Brazil 460 | https://en.wikipedia.org/wiki/France 461 | 271 462 | France 463 | https://en.wikipedia.org/wiki/Italy 464 | 279 465 | Italy 466 | https://en.wikipedia.org/wiki/New_Zealand 467 | 286 468 | New Zealand 469 | https://en.wikipedia.org/wiki/Norway 470 | 299 471 | Norway 472 | https://en.wikipedia.org/wiki/Switzerland 473 | 307 474 | Switzerland 475 | https://en.wikipedia.org/wiki/United_Kingdom 476 | 324 477 | UK 478 | https://en.wikipedia.org/wiki/United_States 479 | 336 480 | US 481 | https://en.wikipedia.org/wiki/United_States 482 | 656 483 | US 484 | https://en.wikipedia.org/wiki/Brazil 485 | 798 486 | Brazil 487 | https://en.wikipedia.org/wiki/Sao_Paulo 488 | 921 489 | Sao Paulo 490 | https://en.wikipedia.org/wiki/Rio_de_Janeiro 491 | 956 492 | Rio de Janeiro 493 | https://en.wikipedia.org/wiki/Brazil 494 | 1091 495 | Brazil 496 | https://en.wikipedia.org/wiki/United_States 497 | 1116 498 | United States 499 | https://en.wikipedia.org/wiki/Brazil 500 | 1228 501 | Brazil 502 | https://en.wikipedia.org/wiki/China 503 | 1345 504 | China 505 | https://en.wikipedia.org/wiki/China 506 | 1709 507 | China 508 | https://en.wikipedia.org/wiki/China 509 | 1912 510 | China 511 | https://en.wikipedia.org/wiki/China 512 | 1978 513 | China 514 | https://en.wikipedia.org/wiki/China 515 | 2207 516 | China 517 | https://en.wikipedia.org/wiki/France 518 | 2228 519 | France 520 | https://en.wikipedia.org/wiki/Kenya 521 | 3615 522 | Kenya 523 | https://en.wikipedia.org/wiki/Kenya 524 | 4276 525 | Kenya 526 | https://en.wikipedia.org/wiki/Nairobi 527 | 4292 528 | Nairobi 529 | https://en.wikipedia.org/wiki/Kisumu 530 | 4401 531 | Kisumu 532 | https://en.wikipedia.org/wiki/Rift_Valley_Province 533 | 4412 534 | Rift Valley 535 | https://en.wikipedia.org/wiki/Mexico 536 | 4499 537 | Mexico 538 | https://en.wikipedia.org/wiki/Mexico 539 | 4784 540 | Mexico 541 | https://en.wikipedia.org/wiki/United_Kingdom 542 | 5040 543 | United Kingdom 544 | https://en.wikipedia.org/wiki/United_Kingdom 545 | 5989 546 | UK 547 | https://en.wikipedia.org/wiki/United_States 548 | 6423 549 | US 550 | https://en.wikipedia.org/wiki/United_States 551 | 6835 552 | United States 553 | https://en.wikipedia.org/wiki/Vietnam 554 | 7066 555 | Vietnam 556 | None 557 | 80 558 | Carancas 559 | https://en.wikipedia.org/wiki/Lake_Titicaca 560 | 95 561 | Lake Titicaca 562 | https://en.wikipedia.org/wiki/Puno_Region 563 | 116 564 | Puno 565 | https://en.wikipedia.org/wiki/Peru 566 | 131 567 | Peru 568 | https://en.wikipedia.org/wiki/Bolivia 569 | 153 570 | Bolivia 571 | https://en.wikipedia.org/wiki/Desaguadero,_Bolivia-Peru 572 | 653 573 | Desaguadero 574 | https://en.wikipedia.org/wiki/Massachusetts 575 | 1474 576 | Massachusetts 577 | https://en.wikipedia.org/wiki/London 578 | 2064 579 | London 580 | https://en.wikipedia.org/wiki/Baghdad 581 | 25 582 | Baghdad 583 | https://en.wikipedia.org/wiki/Iraq 584 | 34 585 | Iraq 586 | https://en.wikipedia.org/wiki/Amman 587 | 321 588 | Amman 589 | https://en.wikipedia.org/wiki/Jordan 590 | 328 591 | Jordan 592 | https://en.wikipedia.org/wiki/United_Kingdom 593 | 414 594 | United Kingdom 595 | https://en.wikipedia.org/wiki/Adhamiyah 596 | 1247 597 | Adhamiya 598 | https://en.wikipedia.org/wiki/Adhamiyah 599 | 1591 600 | Adhamiyah 601 | https://en.wikipedia.org/wiki/Baghdad 602 | 1613 603 | Baghdad 604 | -------------------------------------------------------------------------------- /data/iaa_check.txt: -------------------------------------------------------------------------------- 1 | https://en.wikipedia.org/wiki/Dorset 2 | 97 3 | Dorset 4 | https://en.wikipedia.org/wiki/England 5 | 105 6 | England 7 | https://en.wikipedia.org/wiki/United_Kingdom 8 | 391 9 | United Kingdom 10 | https://en.wikipedia.org/wiki/Michigan 11 | 99 12 | Michigan 13 | https://en.wikipedia.org/wiki/Lake_Erie 14 | 124 15 | Lake Erie 16 | https://en.wikipedia.org/wiki/Pointe_Mouillee_State_Game_Area 17 | 143 18 | Mouillee 19 | https://en.wikipedia.org/wiki/Monroe_County,_Michigan 20 | 171 21 | Monroe County 22 | https://en.wikipedia.org/wiki/Ames,_Iowa 23 | 398 24 | Ames 25 | https://en.wikipedia.org/wiki/Iowa 26 | 404 27 | Iowa 28 | https://en.wikipedia.org/wiki/North_America 29 | 575 30 | North America 31 | https://en.wikipedia.org/wiki/Asia 32 | 676 33 | Asia 34 | https://en.wikipedia.org/wiki/Michigan 35 | 922 36 | Michigan 37 | https://en.wikipedia.org/wiki/Buckinghamshire 38 | 86 39 | Buckinghamshire 40 | https://en.wikipedia.org/wiki/England 41 | 103 42 | England 43 | https://en.wikipedia.org/wiki/Milton_Keynes 44 | 334 45 | Milton Keynes 46 | https://en.wikipedia.org/wiki/Peru 47 | 25 48 | Peru 49 | https://en.wikipedia.org/wiki/Peru 50 | 175 51 | Peru 52 | https://en.wikipedia.org/wiki/Peru 53 | 339 54 | Peru 55 | https://en.wikipedia.org/wiki/Peru 56 | 516 57 | Peru 58 | https://en.wikipedia.org/wiki/Libya 59 | 1 60 | Libya 61 | https://en.wikipedia.org/wiki/Sofia 62 | 224 63 | Sofia 64 | https://en.wikipedia.org/wiki/Bulgaria 65 | 231 66 | Bulgaria 67 | https://en.wikipedia.org/wiki/Bulgaria 68 | 265 69 | Bulgaria 70 | https://en.wikipedia.org/wiki/Benghazi 71 | 548 72 | Benghazi 73 | https://en.wikipedia.org/wiki/Libya 74 | 648 75 | Libya 76 | https://en.wikipedia.org/wiki/Libya 77 | 800 78 | Libya 79 | https://en.wikipedia.org/wiki/Libya 80 | 912 81 | Libya 82 | https://en.wikipedia.org/wiki/Libya 83 | 993 84 | Libya 85 | https://en.wikipedia.org/wiki/Qatar 86 | 931 87 | Qatar 88 | https://en.wikipedia.org/wiki/Benghazi 89 | 1136 90 | Benghazi 91 | https://en.wikipedia.org/wiki/Libya 92 | 1276 93 | Libya 94 | https://en.wikipedia.org/wiki/Libya 95 | 1369 96 | Libya 97 | https://en.wikipedia.org/wiki/Libya 98 | 1395 99 | Libya 100 | https://en.wikipedia.org/wiki/France 101 | 1558 102 | France 103 | https://en.wikipedia.org/wiki/France 104 | 1614 105 | France 106 | https://en.wikipedia.org/wiki/Libya 107 | 1635 108 | Libya 109 | https://en.wikipedia.org/wiki/Libya 110 | 1670 111 | Libya 112 | https://en.wikipedia.org/wiki/Tripoli 113 | 1697 114 | Tripoli 115 | https://en.wikipedia.org/wiki/Europe 116 | 1770 117 | Europe 118 | https://en.wikipedia.org/wiki/Paris 119 | 1875 120 | Paris 121 | https://en.wikipedia.org/wiki/Bulgaria 122 | 2175 123 | Bulgaria 124 | https://en.wikipedia.org/wiki/London 125 | 81 126 | London 127 | https://en.wikipedia.org/wiki/Liverpool 128 | 178 129 | Liverpool 130 | https://en.wikipedia.org/wiki/United_Kingdom 131 | 613 132 | UK 133 | https://en.wikipedia.org/wiki/United_Kingdom 134 | 739 135 | UK 136 | https://en.wikipedia.org/wiki/Haiti 137 | 424 138 | Haiti 139 | https://en.wikipedia.org/wiki/Central_Plateau_(Haiti) 140 | 711 141 | Central Plateau 142 | https://en.wikipedia.org/wiki/Artibonite_(department) 143 | 731 144 | Artibonite 145 | https://en.wikipedia.org/wiki/Port-au-Prince 146 | 804 147 | Port-Au-Prince 148 | https://en.wikipedia.org/wiki/M%C3%B4le-Saint-Nicolas 149 | 970 150 | St. Nicolas 151 | https://en.wikipedia.org/wiki/Saint-Marc 152 | 1030 153 | St. Marc 154 | https://en.wikipedia.org/wiki/Java 155 | 81 156 | Java 157 | https://en.wikipedia.org/wiki/Java 158 | 341 159 | Java 160 | https://en.wikipedia.org/wiki/Pangandaran 161 | 936 162 | Pangandaran 163 | https://en.wikipedia.org/wiki/Indonesia 164 | 1227 165 | Indonesia 166 | https://en.wikipedia.org/wiki/Pangandaran 167 | 1282 168 | Pangandaran 169 | https://en.wikipedia.org/wiki/United_States 170 | 1519 171 | US 172 | https://en.wikipedia.org/wiki/Kitaakita 173 | 90 174 | Kitaakita 175 | https://en.wikipedia.org/wiki/Akita_Prefecture 176 | 115 177 | Akita Prefecture 178 | https://en.wikipedia.org/wiki/Japan 179 | 144 180 | Japan 181 | https://en.wikipedia.org/wiki/United_States 182 | 92 183 | U.S. 184 | https://en.wikipedia.org/wiki/United_States 185 | 225 186 | United States 187 | https://en.wikipedia.org/wiki/United_States 188 | 288 189 | United States 190 | https://en.wikipedia.org/wiki/New_York_City 191 | 1202 192 | New York City 193 | https://en.wikipedia.org/wiki/Queens 194 | 1306 195 | Queens 196 | https://en.wikipedia.org/wiki/New_York_City 197 | 1314 198 | New York 199 | https://en.wikipedia.org/wiki/Ohio 200 | 1495 201 | Ohio 202 | https://en.wikipedia.org/wiki/California 203 | 1572 204 | California 205 | https://en.wikipedia.org/wiki/Kansas 206 | 1634 207 | Kansas 208 | https://en.wikipedia.org/wiki/Texas 209 | 1645 210 | Texas 211 | https://en.wikipedia.org/wiki/United_States 212 | 1923 213 | U.S. 214 | https://en.wikipedia.org/wiki/United_States 215 | 2090 216 | U.S. 217 | https://en.wikipedia.org/wiki/Kabanjahe 218 | 72 219 | Kabanjahe 220 | https://en.wikipedia.org/wiki/Sumatra 221 | 94 222 | Sumatra 223 | https://en.wikipedia.org/wiki/Indonesia 224 | 105 225 | Indonesia 226 | https://en.wikipedia.org/wiki/Kabanjahe 227 | 308 228 | Kabanjahe 229 | https://en.wikipedia.org/wiki/Indonesia 230 | 351 231 | Indonesia 232 | https://en.wikipedia.org/wiki/Atlanta 233 | 501 234 | Atlanta 235 | https://en.wikipedia.org/wiki/Georgia_(U.S._state) 236 | 510 237 | Georgia 238 | https://en.wikipedia.org/wiki/Atlanta 239 | 670 240 | Atlanta 241 | https://en.wikipedia.org/wiki/United_States 242 | 704 243 | U.S. 244 | https://en.wikipedia.org/wiki/China 245 | 85 246 | China 247 | https://en.wikipedia.org/wiki/Beijing 248 | 121 249 | Beijing 250 | https://en.wikipedia.org/wiki/China 251 | 401 252 | China 253 | https://en.wikipedia.org/wiki/Hebei 254 | 658 255 | Hebei 256 | https://en.wikipedia.org/wiki/Hubei 257 | 787 258 | Hubei 259 | https://en.wikipedia.org/wiki/China 260 | 871 261 | People's Republic of China 262 | https://en.wikipedia.org/wiki/China 263 | 991 264 | China 265 | https://en.wikipedia.org/wiki/Anhui 266 | 1144 267 | Anhui 268 | https://en.wikipedia.org/wiki/Fuyang 269 | 1277 270 | Fuyang 271 | https://en.wikipedia.org/wiki/Anhui 272 | 1287 273 | Anhui 274 | https://en.wikipedia.org/wiki/Beijing 275 | 1532 276 | Beijing 277 | https://en.wikipedia.org/wiki/Guangdong 278 | 1589 279 | Guangdong 280 | https://en.wikipedia.org/wiki/Guangxi 281 | 1600 282 | Guangxi 283 | https://en.wikipedia.org/wiki/Hainan 284 | 1609 285 | Hainan 286 | https://en.wikipedia.org/wiki/Hunan 287 | 1617 288 | Hunan 289 | https://en.wikipedia.org/wiki/Zhejiang 290 | 1624 291 | Zhejiang 292 | https://en.wikipedia.org/wiki/Hubei 293 | 1646 294 | Hubei 295 | https://en.wikipedia.org/wiki/China 296 | 1853 297 | China 298 | https://en.wikipedia.org/wiki/China 299 | 1982 300 | China 301 | https://en.wikipedia.org/wiki/Beijing 302 | 2196 303 | Beijing 304 | https://en.wikipedia.org/wiki/China 305 | 2302 306 | China 307 | https://en.wikipedia.org/wiki/China 308 | 2394 309 | China 310 | https://en.wikipedia.org/wiki/China 311 | 2458 312 | China 313 | https://en.wikipedia.org/wiki/Asia 314 | 81 315 | Asia 316 | https://en.wikipedia.org/wiki/India 317 | 113 318 | India 319 | https://en.wikipedia.org/wiki/Nepal 320 | 120 321 | Nepal 322 | https://en.wikipedia.org/wiki/Bangladesh 323 | 131 324 | Bangladesh 325 | https://en.wikipedia.org/wiki/Bihar 326 | 462 327 | Bihar 328 | https://en.wikipedia.org/wiki/Uttar_Pradesh 329 | 472 330 | Uttar Pradesh 331 | https://en.wikipedia.org/wiki/Bihar 332 | 586 333 | Bihar 334 | https://en.wikipedia.org/wiki/India 335 | 795 336 | India 337 | None 338 | 1113 339 | Gondra 340 | https://en.wikipedia.org/wiki/Uttar_Pradesh 341 | 1132 342 | Uttar Pradesh 343 | https://en.wikipedia.org/wiki/Indore 344 | 1280 345 | Indore 346 | https://en.wikipedia.org/wiki/Greenlane 347 | 77 348 | Greenlane 349 | https://en.wikipedia.org/wiki/Auckland 350 | 88 351 | Auckland 352 | https://en.wikipedia.org/wiki/New_Zealand 353 | 98 354 | New Zealand 355 | https://en.wikipedia.org/wiki/Auckland 356 | 253 357 | Auckland 358 | https://en.wikipedia.org/wiki/Greenlane 359 | 1570 360 | Greenlane 361 | https://en.wikipedia.org/wiki/Auckland 362 | 1581 363 | Auckland 364 | https://en.wikipedia.org/wiki/Auckland 365 | 1680 366 | Auckland 367 | https://en.wikipedia.org/wiki/Zimbabwe 368 | 46 369 | Zimbabwe 370 | https://en.wikipedia.org/wiki/Zimbabwe 371 | 294 372 | Zimbabwe 373 | https://en.wikipedia.org/wiki/Zimbabwe 374 | 388 375 | Zimbabwe 376 | https://en.wikipedia.org/wiki/Zimbabwe 377 | 712 378 | Zimbabwe 379 | https://en.wikipedia.org/wiki/Harare 380 | 788 381 | Harare 382 | https://en.wikipedia.org/wiki/Mabvuku 383 | 898 384 | Mabvuku 385 | https://en.wikipedia.org/wiki/Harare 386 | 932 387 | Harare 388 | https://en.wikipedia.org/wiki/Botswana 389 | 1186 390 | Botswana 391 | https://en.wikipedia.org/wiki/Mozambique 392 | 1196 393 | Mozambique 394 | https://en.wikipedia.org/wiki/Zimbabwe 395 | 1237 396 | Zimbabwe 397 | https://en.wikipedia.org/wiki/South_Africa 398 | 1212 399 | South Africa 400 | https://en.wikipedia.org/wiki/Zimbabwe 401 | 1519 402 | Zimbabwe 403 | https://en.wikipedia.org/wiki/Zimbabwe 404 | 1602 405 | Zimbabwe 406 | https://en.wikipedia.org/wiki/Zimbabwe 407 | 4 408 | Zimbabwe 409 | https://en.wikipedia.org/wiki/Wedza_District 410 | 55 411 | Wedza 412 | https://en.wikipedia.org/wiki/Mashonaland_East_Province 413 | 71 414 | Mashonaland East 415 | https://en.wikipedia.org/wiki/Midlands_Province 416 | 550 417 | Midlands 418 | https://en.wikipedia.org/wiki/Zimbabwe 419 | 1242 420 | Zimbabwe 421 | https://en.wikipedia.org/wiki/United_States 422 | 27 423 | United States 424 | https://en.wikipedia.org/wiki/United_States 425 | 294 426 | United States 427 | https://en.wikipedia.org/wiki/Iraq 428 | 439 429 | Iraq 430 | https://en.wikipedia.org/wiki/Afghanistan 431 | 445 432 | Afghanistan 433 | https://en.wikipedia.org/wiki/North_Korea 434 | 458 435 | North Korea 436 | https://en.wikipedia.org/wiki/Iran 437 | 559 438 | Iran 439 | https://en.wikipedia.org/wiki/United_States 440 | 916 441 | U.S. 442 | https://en.wikipedia.org/wiki/Mexico 443 | 923 444 | Mexico 445 | https://en.wikipedia.org/wiki/China 446 | 170 447 | China 448 | https://en.wikipedia.org/wiki/Indonesia 449 | 177 450 | Indonesia 451 | https://en.wikipedia.org/wiki/Africa 452 | 209 453 | Africa 454 | https://en.wikipedia.org/wiki/China 455 | 493 456 | China 457 | https://en.wikipedia.org/wiki/Indonesia 458 | 503 459 | Indonesia 460 | https://en.wikipedia.org/wiki/Africa 461 | 517 462 | Africa 463 | https://en.wikipedia.org/wiki/Rome 464 | 720 465 | Rome 466 | https://en.wikipedia.org/wiki/Italy 467 | 726 468 | Italy 469 | https://en.wikipedia.org/wiki/Africa 470 | 1118 471 | Africa 472 | https://en.wikipedia.org/wiki/Indonesia 473 | 1392 474 | Indonesia 475 | https://en.wikipedia.org/wiki/Java 476 | 1716 477 | Java 478 | https://en.wikipedia.org/wiki/China 479 | 1814 480 | China 481 | https://en.wikipedia.org/wiki/China 482 | 1859 483 | China 484 | https://en.wikipedia.org/wiki/Mexico 485 | 48 486 | Mexico 487 | https://en.wikipedia.org/wiki/Australia 488 | 253 489 | Australia 490 | https://en.wikipedia.org/wiki/Brazil 491 | 264 492 | Brazil 493 | https://en.wikipedia.org/wiki/France 494 | 272 495 | France 496 | https://en.wikipedia.org/wiki/Italy 497 | 280 498 | Italy 499 | https://en.wikipedia.org/wiki/New_Zealand 500 | 287 501 | New Zealand 502 | https://en.wikipedia.org/wiki/Norway 503 | 300 504 | Norway 505 | https://en.wikipedia.org/wiki/Switzerland 506 | 308 507 | Switzerland 508 | https://en.wikipedia.org/wiki/United_Kingdom 509 | 325 510 | UK 511 | https://en.wikipedia.org/wiki/United_States 512 | 337 513 | US 514 | https://en.wikipedia.org/wiki/United_States 515 | 657 516 | US 517 | https://en.wikipedia.org/wiki/Brazil 518 | 799 519 | Brazil 520 | https://en.wikipedia.org/wiki/Sao_Paulo 521 | 922 522 | Sao Paulo 523 | https://en.wikipedia.org/wiki/Rio_de_Janeiro 524 | 957 525 | Rio de Janeiro 526 | https://en.wikipedia.org/wiki/Brazil 527 | 1092 528 | Brazil 529 | https://en.wikipedia.org/wiki/United_States 530 | 1117 531 | United States 532 | https://en.wikipedia.org/wiki/Argentina 533 | 1218 534 | Argentina 535 | https://en.wikipedia.org/wiki/Brazil 536 | 1229 537 | Brazil 538 | https://en.wikipedia.org/wiki/China 539 | 1346 540 | China 541 | https://en.wikipedia.org/wiki/China 542 | 1710 543 | China 544 | https://en.wikipedia.org/wiki/China 545 | 1913 546 | China 547 | https://en.wikipedia.org/wiki/China 548 | 1979 549 | China 550 | https://en.wikipedia.org/wiki/China 551 | 2208 552 | China 553 | https://en.wikipedia.org/wiki/France 554 | 2229 555 | France 556 | https://en.wikipedia.org/wiki/India 557 | 2723 558 | India 559 | https://en.wikipedia.org/wiki/India 560 | 2856 561 | India 562 | https://en.wikipedia.org/wiki/Kenya 563 | 3616 564 | Kenya 565 | https://en.wikipedia.org/wiki/Kenya 566 | 4277 567 | Kenya 568 | https://en.wikipedia.org/wiki/Nairobi 569 | 4293 570 | Nairobi 571 | https://en.wikipedia.org/wiki/Kisumu 572 | 4402 573 | Kisumu 574 | https://en.wikipedia.org/wiki/Rift_Valley_Province 575 | 4413 576 | Rift Valley 577 | https://en.wikipedia.org/wiki/Mexico 578 | 4500 579 | Mexico 580 | https://en.wikipedia.org/wiki/Mexico 581 | 4785 582 | Mexico 583 | https://en.wikipedia.org/wiki/United_Kingdom 584 | 5041 585 | United Kingdom 586 | https://en.wikipedia.org/wiki/United_Kingdom 587 | 5990 588 | UK 589 | https://en.wikipedia.org/wiki/United_Kingdom 590 | 6139 591 | Britain 592 | https://en.wikipedia.org/wiki/United_States 593 | 6424 594 | US 595 | https://en.wikipedia.org/wiki/United_States 596 | 6836 597 | United States 598 | https://en.wikipedia.org/wiki/United_States 599 | 7021 600 | US 601 | https://en.wikipedia.org/wiki/Vietnam 602 | 7067 603 | Vietnam 604 | https://en.wikipedia.org/wiki/Lake_Titicaca 605 | 96 606 | Lake Titicaca 607 | https://en.wikipedia.org/wiki/Puno_Region 608 | 117 609 | Puno 610 | https://en.wikipedia.org/wiki/Peru 611 | 132 612 | Peru 613 | https://en.wikipedia.org/wiki/Bolivia 614 | 154 615 | Bolivia 616 | https://en.wikipedia.org/wiki/Desaguadero,_Bolivia-Peru 617 | 654 618 | Desaguadero 619 | https://en.wikipedia.org/wiki/Massachusetts 620 | 1475 621 | Massachusetts 622 | https://en.wikipedia.org/wiki/London 623 | 2065 624 | London 625 | https://en.wikipedia.org/wiki/Baghdad 626 | 26 627 | Baghdad 628 | https://en.wikipedia.org/wiki/Iraq 629 | 35 630 | Iraq 631 | https://en.wikipedia.org/wiki/Amman 632 | 322 633 | Amman 634 | https://en.wikipedia.org/wiki/Jordan 635 | 329 636 | Jordan 637 | https://en.wikipedia.org/wiki/United_Kingdom 638 | 415 639 | United Kingdom 640 | https://en.wikipedia.org/wiki/Adhamiyah 641 | 1248 642 | Adhamiya 643 | https://en.wikipedia.org/wiki/Adhamiyah 644 | 1592 645 | Adhamiyah 646 | https://en.wikipedia.org/wiki/Baghdad 647 | 1614 648 | Baghdad 649 | -------------------------------------------------------------------------------- /data/iaa_test.txt: -------------------------------------------------------------------------------- 1 | -------------NEW ARTICLE----------------- 2 | The H5N1 Avian Flu virus has been found in a dead wild Canadian Goose in Abbotsbury Swannery in Dorset, England. This is the eleventh case of the virus turning up in wild birds. The goose was discovered on February 25, 2008. "The finding of more cases in wild birds is not unexpected ... We are currently considering whether any additional restrictions are necessary in the area," said the United Kingdom's Health Ministry in a statement to the media. As a result of the finding, poultry movement has now been restricted by the Department for the Environment, Food and Rural Affairs (DEFRA) in the areas surrounding the swannery. The removal of all birds, dead or alive from any property now requires a license. DEFRA says the restrictions will expire no earlier than 31 days. 3 | -------------NEW ARTICLE----------------- 4 | Scientists have discovered the possible presence of the H5N1 Bird Flu virus in wild mute swans in Michigan on the coast of Lake Erie near the Mouillee state game area in Monroe County. The swans were sampled on August 8, 2006 and the initial testing was done at Michigan State University's Diagnostic Center for Population and Animal Health and at the National Veterinary Services laboratories in Ames, Iowa. White House Press Secretary Tony Snow says that "They (the scientists) believe it is a strain of low pathogenicity, similar to strains that have been seen before in North America." Snow also added that "this [case] is not what we're accustomed to hearing about from Asia." "Test results thus far indicate this is low pathogenicity avian influenza, which poses no threat to human health. Routine surveillance has indicated the presence of H5 and N1 avian influenza subtypes in samples from two wild mute swans in Michigan," said a statement on the United States Department of Agriculture's (USDA) website. The statement also went on to say that the swans "did not show signs of sickness" and that the swans were infected with "two separate" strains of Avian Flu. "It is possible that these birds were not infected with an H5N1 strain, but instead with two separate avian influenza viruses, one containing H5 and the other containing N1," said the statement. "This is not the highly pathogenic avian influenza virus that has spread through much of other parts of the world. We do not believe this virus represents a risk to human health," said USDA's Animal and Plant Health inspector, Ron DeHaven. Health officials are remaining "remaining vigilant and prepared," said Department of Health and Human Services science advisor, Dr. William Raub. Further tests will be done to confirm that there is in fact a virus there and what type and are expected sometime today. 5 | -------------NEW ARTICLE----------------- 6 | A five-year-old boy has been suspected to have died from the H1N1 swine flu virus in Buckinghamshire, England. The boy came from Emberton School, which now has just 29 pupils attending. Health tests are currently being carried out to determine whether or not the child did indeed die from the virus. He was admitted to a hospital in Milton Keynes, but later died in the early hours of Sunday morning. At present, the individual remains unidentified. Steve Dunning is the principal in the school. Speaking to BBC Three Counties Radio, Dunning said: "The staff of Emberton School are very saddened to learn of the death of one of their pupils who was a confident, delightful and happy student and will be missed greatly. At this time we are focusing on supporting the children and parents in our small village community. I have spoken directly with the mother and passed on the condolences of all the staff and governors at the school." 7 | -------------NEW ARTICLE----------------- 8 | At least four people in Peru have contracted HIV, the virus that causes AIDS after receiving blood transfusions that were infected with the deadly virus. As a result, all of Peru's 240 blood banks have been closed pending further investigation and an emergency has been declared by the government. "This situation cannot continue. All of Peru's blood banks are being reviewed. We do not want people to panic, what we have to do is be more careful, strengthen our care," said Carlos Vallejos, the Health Minister of Peru. The country's government made its decision when Judith Rivera, 44 and mother of four children, went public with claims that during an operation, she was infected with HIV when doctors gave her a transfusion. Soon after the government admitted that at least three other individuals, one being an infant child only 11 months old, another a 17 year-old boy, contracted the virus through blood transfusions and all received care from the same hospitals. Vallejos also stated that all of the individuals infected will receive whatever care is necessary to treat their conditions. 9 | -------------NEW ARTICLE----------------- 10 | Libya has freed six foreign medical personnel who were convicted of infecting hundreds of Libyan children with HIV and sentenced to death. In jail since 1999, the five Bulgarian nurses and one Palestinian doctor arrived in Sofia, Bulgaria, today. The president of Bulgaria, Georgi Parvanov, promptly pardoned them. All six have maintained their innocence throughout. They have also claimed that they suffered torture to extract confessions. International HIV experts testified at the trials that the infections began before the six arrived at the Benghazi hospital. They said the infections were more likely the result of poor hygiene. Last week, Libya lifted the death sentences following a US$460 million financial settlement, which works out to US$1 million to each HIV victim's family. However, Libya insisted on further concessions on relations with the European Union and aid. A deal between the E.U. and Libya, mediated by Qatar, ended the diplomatic standoff. The foreign minister of Libya, Abdel Rahman Shalgham, said the E.U. promised to provide "life-long treatment" to the infected children, as well as aid to "improve the Benghazi hospital" where the children were infected. Further, he claimed that deal will allow for "full cooperation and partnership between Libya and the European Union." "We hope to go on further [in] normalising our relations with Libya. Our relations with Libya were to a large extent blocked by the non-settlement of this medics issue," said José Manuel Barroso, President of the European Commission. The president of France, Nicolas Sarkozy, said that neither the E.U. nor France paid money to Libya. He also said he would visit Libya on Wednesday to help Tripoli's reintegration. "I can quite simply confirm to you that neither Europe nor France have made the slightest financial contribution to Libya," said Sarkozy to reporters in Paris. "I have had the opportunity to thank the Qatari authorities very warmly for their mediation and their humanitarian intervention." Benita Ferrero-Waldner, the European Commissioner for External Relations, said: "I share the joy of their families and friends and of the government and people of Bulgaria. For over eight years, we have never forgotten the suffering of the medical staff who have shown such dignity and fortitude during their long ordeal." "Now I still can't believe that I am standing on Bulgarian soil. We were told the news at four o'clock in the morning and we left the jail at quarter to six to board the plane. Now I will try to get my previous life back," said Kristiyana Vulcheva, 48, upon her release at the airport. 11 | -------------NEW ARTICLE----------------- 12 | Mr Justice Saunders today imprisoned a man for eight years at the Old Bailey in London after an FBI sting in which he tried to buy ricin on the Dark Web. Mohammed Ali, 31, from Liverpool, was convicted at trial of attempting to possess a chemical weapon. He told the jury he was "curious" about the Dark Web, which is a largely hidden and difficult to police section of the Internet. Ali said he didn't realise he had done anything illegal. Ali was prosecuted under the Chemical Weapons Act 1996 after sending an undercover agent a message reading "Hi, would you be able to make me some ricin and send it to the UK?" He bought 500mg, which has the potential to kill about 1,400, but was sent a dummy package. Counter-terror police in the UK liaised with the FBI. The powder, which Ali paid for in BitCoin, was actually harmless. Hidden inside a toy car, the package was treated with markers and his face glowed under ultraviolet light, indicating he had handled it. The judge said today there was "real risk" involved. Ali told his trial he had discovered drugs and guns for sale. Computer analysis showed he had looked up ricin and other poisons; he said he went for ricin simply because it had appeared in Breaking Bad. He also searched for small animals after being advised to test it on a rodent; a to-do list on his computer included "paid ricin guy" and "get pet to murder". 13 | -------------NEW ARTICLE----------------- 14 | Nearly 200 people are confirmed dead and approximately 2600 are ill in a central Haitian cholera outbreak, according to the Centers for Disease Control (CDC), the U.S. Agency for International Development (USAID) and United Nations (UN). Haitian officials place the death toll at 194 deaths with 2,364 people being infected. According to CDC officials Dr. Rob Quick and Dr. Carleene Dei an eleven man team is being sent to Haiti to investigate and determine the best course of action for the country. The USAID has said that they will provide supplies to set up treatment centers and have already provided 300,000 oral re-hydration kits and water purification kits. A majority of the reported cases are in the Central Plateau and Artibonite regions located just north of the earthquake ravaged capital, Port-Au-Prince. Officials fear that the disease could spread to the capital city if not brought under control. It has been reported that many people are flooding the St. Nicolas hospital, which is the main medical facility in St. Marc, for treatment, causing pandemonium outside the gate. An aid worker who visited the hospital called it a "horror scene", while another worker, David Darg, of Operation Blessing International wrote "The courtyard was lined with patients hooked up to intravenous drips. It had just rained and there were people lying on the ground on soggy sheets, half-soaked with feces." 15 | -------------NEW ARTICLE----------------- 16 | More than 650 people have now died after a tsunami hit the Indonesian island of Java on Monday afternoon. In the past few days, around 100 dead bodies have been recovered, and it is estimated that over 300 people are still missing. An underwater earthquake with a magnitude of 7.7 triggered the deadly wave which ravaged a 200km stretch of Java's southern coast. Thousands of people are continuing to camp in the hills. They are too apprehensive to return home due to fears of another tsunami, but according to Reuters, health officials are worried about the threat of disease among those who are still in refuge. "The risk of catching diseases is there because they live in an open area with limited tents and water," said Rustan Pakaya, from the health ministry's crisis centre. He added that people were being given injections to protect them from diseases like measles, tetanus and cholera. Areas worst hit, like the small town of Pangandaran, are beginning to return to normal, and many businesses there have begun to open up again. "The market and many shops are already open today and although they are not operating fully, things are slowly returning to normal," district spokesman Wasdi bin Umri told AFP. Yesterday, Indonesia's President, Susilo Bambang Yudhoyono toured Pangandaran and met people who were staying in a temporary camp. The Indonesian government has been criticised for failing to inform residents living on the coast that a tsunami was looming. After the underwater earthquake was detected, US and Japanese agencies issued warning notices, but the government has admitted it was unable to transmit the bulletins to coastal areas. Speaking yesterday, Mr Yudhoyono vowed to hasten efforts to build an early warning system planned after the 2004 Asian tsunami. "We want to expedite efforts to get infrastructure for the tsunami warning system in place," AP quoted him as saying. "I will work with parliament to get the budget". 17 | -------------NEW ARTICLE----------------- 18 | Six people have died after becoming infected with the Hong Kong Flu virus in the city of Kitaakita located in the Akita Prefecture of northern Japan. Among the dead were four men and two women, aged 60 to 90 years old. All of them died between Monday November 1, and Friday November 5, while in a local hospital. Dozens more are suspected to be ill with the virus while another 30 others, mostly doctors and other hospital personnel, are being tested for the virus. Preliminary tests on the patients showed a positive result for the infection while further laboratory tests confirmed the presence of it. The virus killed nearly 1,000,000 people around the globe in 1968 and 1969. It was a form of the H3N2 strain of influenza type A. So far there are no reports of anyone infected outside hospital grounds. 19 | -------------NEW ARTICLE----------------- 20 | Homeland Security Secretary Janet Napolitano announced today at a news conference that the U.S. has declared a public health emergency in light of the swine flu outbreak. The total number of confirmed swine flu cases in the United States stands at 20. Secretary Napolitano said that the United States' declaration follows suit with the "standard operating procedure" of such an outbreak to make more government resources available to combat the disease. One direct result of the declaration is the government's mobilization of approximately 12 million doses of Tamiflu to locations where the states can quickly access their share of the medication if needed. Secretary Napolitano urged residents not to panic saying that the government is issuing a "declaration of emergency preparedness." Secretary Napolitano added, "Really that's what we're doing right now. We're preparing in an environment where we really don't know ultimately what the size of seriousness of this outbreak is going to be." John Brennan, a Homeland Security assistant, added that "at this point, a top priority is to ensure that communication is robust and that medical surveillance efforts are fully activated." This afternoon, New York City Mayor Michael Bloomberg reported that 8 students from the St. Francis Prepatory School in Queens, New York have contracted the swine flu. All in all, more than 100 students from that high school were absent last week with flu-like symptoms. Meanwhile, public health officials in Ohio today announced one confirmed case of swine flu in the state. Thus far, California has reported 7 confirmed cases of swine flu, while Kansas and Texas have each reported two confirmed cases. At the same news conference Dr. Richard Besser, the acting director from the Centers for Disease Control and Prevention (CDC), said to expect additional cases of swine flu to be reported in the short term. Dr. Besser added that the U.S. could also start seeing cases of the disease where the effects are more dramatic: "We're going to see more severe disease in this country". So far, no one in the U.S. has died from swine flu. 21 | -------------NEW ARTICLE----------------- 22 | Preliminary tests performed on samples taken from six villagers in the Kabanjahe district of Sumatra in Indonesia have tested negative for the deadly H5N1 Avian Flu virus. "Investigations by the ministry of health lab and Namru, too, on August 2 and 3 on all specimens collected from the suspected cases in Kabanjahe district came up negative," said Indonesia's health minister, Siti Fadilah Supari. Final test results are expected in at least seven days from the Center for Disease Control (CDC) in Atlanta, Georgia. "The World Health Organization (WHO) requires human samples to be sent to one of WHO's six collaborative centres. So, we only need to send them to CDC Atlanta as it has worked with the U.S. NAMRU-2 lab here," added Supari. Supari also stated that all individuals are suffering from the "common flu." 23 | -------------NEW ARTICLE----------------- 24 | Over 40 children have died in an outbreak of hand, foot and mouth disease (HFMD) in China, and the country's capital of Beijing reported its first death due to the disease on Wednesday. According to Xinhua News Agency, Beijing Health Bureau spokeswoman Deng Xiaohong said that the 13-month-old boy died Sunday while en route to the hospital. Health authorities state that 24,934 children in mainland China are afflicted with the disease, and 42 children have died from it. The cause of the disease has been identified as Enterovirus 71 (EV-71). HFMD can also be caused by Coxsackievirus. Another child infected with the virus died Monday, but as he died in Hebei province his death was counted there. Xinhua News Agency also reported that a 21-month-old boy died Monday of the virus in Hubei province. After an order was given last week by the Ministry of Health of the People's Republic of China that all cases must be reported, the count of those infected has increased markedly. Eastern China saw a large number of cases in early March, but this information was not made public until late April. In March, Children under age six in eastern Anhui province began being admitted to hospitals with symptoms of the virus, and the outbreak spread quickly after that. The city of Fuyang in Anhui province was especially hard-hit by the outbreak. "The majority of patients who were in critical condition have recovered," said a Health Ministry official in a statement on Monday. As of Monday, 3,606 HFMD infections had been reported in Beijing. Deaths have occurred in the provinces of Anhui, Guangdong, Guangxi, Hainan, Hunan, Zhejiang, Beijing and Hubei. "What I know is the death rate has gone down drastically since early May. There are very, very few cases with complications — 99 percent of these are mild cases," said World Health Organization (WHO) China representative Hans Troedsson in a statement on Wednesday. Incidents of the disease are expected to peak in June and July. China is also dealing with a magnitude-7.9 earthquake which hit the country Monday and has killed almost 15,000. The outbreak is a concern to the government, as the country prepares for the 2008 Summer Olympics in Beijing this August. "We are confident the potential outbreak will not affect the Beijing Olympic Games," China's Health Ministry spokesman Mao Qunan stated. And at a joint press conference held by China's Ministry of Health and the WHO, he further noted that, "China is confident that it can control the spread of the disease with effective prevention methods." 25 | -------------NEW ARTICLE----------------- 26 | Torrential monsoon rains for the past 11 days have inundated parts of southeast Asia, heavily flooding areas of India, Nepal, and Bangladesh and affecting over 20 million people. An estimated 1,400 have been killed across the region, as waterborne diseases have proliferated and become highly virulent in the humid and wet conditions. The United Nations Children's Fund, UNICEF, is currently providing relief to affected areas. In the Northern Indian states of Bihar and Uttar Pradesh, as many as 12.5 million people have been displaced or affected by flooding. Clashes with police in Bihar have left 20 villagers injured and one man dead, since many are staging protests after being forced from their homes. UNICEF estimates as many as 1,103 may have died in 138 districts throughout northern India. Hospitals are reaching dangerous volumes as disease is sickening thousands. Many are ill from dehydration, exposure and dysentery. Aid workers warn that malaria is a serious risk. The Indian Air Force began dropping aid packages containing medical supplies and food yesterday. Santosh Mishra, a villager in the Gondra district of Uttar Pradesh remarked, "I have not seen such flooding in the last 24 years." Bodies of two students of the IIM Indore swept in the flash floods in Indore were recovered. 27 | -------------NEW ARTICLE----------------- 28 | McDonald's is asking everyone who ate at a McDonald's restaurant located in Greenlane, Auckland, New Zealand, in December 15, 2006, to go to their doctor after a worker tested positive for Hepatitis A. Doctor Greg Simmons, medical officer of health in Auckland, said that it is possible for the food handler worker to have passed on the disease during 7:00 p.m. (NZDT) and 2:00 a.m. as he would have been in the most infectious stage of Hepatitis A and was not wearing gloves. According to a spokeswoman from McDonald's, Joanna Redfern-Hardisty, food handlers are not required to wear gloves but are required to have thoroughly washed their hands with antimicrobial soap. However Dr Simmons says that the infected worker usually wore gloves. The shift the worker worked was the only one the worker handled food when he was infectious with Hepatitis A. It is unknown how many people could have come into contact with food that the infected worker had prepared as Friday night is usually a busy night, according to Ms Redfern-Hardisty. Dr Simmons says that if someone has been infected with Hepatitis A then they will currently be showing early symptoms. Those symptoms include being tired, no appetite, nausea and skin and eyes will show a yellowish colour. Ms Redfern-Hardisty, spokeswoman for McDonald's, confirmed that no more risk is being posed to customers. She said that it is currently unknown where the worker caught Hepatitis A but has stated that it was from an external source. The worker has been suspended from work. If you ate at the McDonalds located in Greenlane, Auckland, on late December 15, early December 16, then please either visit your doctor or ring the Auckland Regional Public Health Service on 096234600. 29 | -------------NEW ARTICLE----------------- 30 | A cholera outbreak in the African country of Zimbabwe has killed almost 500 people since August, according to the World Health Organization (WHO). The WHO said that the outbreak affected most areas of the country, and that some remote areas had seen fatality rates increase by as much as 30%. Zimbabwe's Ministry of Health reported 484 deaths from 11,735 cases since the outbreak began. Zimbabwe has had annual outbreaks of cholera for nearly a decade, but this one was the most far-reaching. A report by the WHO stated that the last large outbreak was in 1992, with 3,000 cases recorded. Cholera is frequently spread by contaminated, untreated water. The spread of the disease was expedited by the collapse of Zimbabwe's health and sanitation systems; state media reported that most of Harare has been left without water after the city ran out of chemicals for its treatment plant. A resident of Mabvuku, a suburb located east of Harare, told APTV that electricity is not available most of the time, so water is consumed without being boiled first. The medical charity Médecins Sans Frontières (Doctors Without Borders) said that there have been cases of cholera reported in areas of Botswana, Mozambique, and South Africa that border Zimbabwe, indicating the sub-regional threat of the outbreak. The South African ministry of health confirmed that they had 160 incidents of cholera reported, as well as three deaths. The European Commission said that it was providing €9 million (US$11.4 million) in funds to assist Zimbabwe with the crisis. "I'm shocked at the deteriorating humanitarian crisis in Zimbabwe and call upon the authorities there to respond quickly to this cholera outbreak by allowing full assistance from international humanitarians and regional partners," said the commissioner responsible for the European Union's humanitarian aid, Louis Michel. Other agencies providing aid to the country include the United Nations Children's Fund, the WHO, and Doctors Without Borders. 31 | -------------NEW ARTICLE----------------- 32 | As Zimbabwe's rainy season approaches, farmers in the Wedza district, Mashonaland East province, are worried that again this year they won't be able to locate or afford the seed, fertilizer, and other inputs needed to get a maize crop in the ground. Aiming to relieve such shortages of inputs, Christian Care is among the non-governmental organizations working with the United Nations Food and Agriculture Organization and the European Union to implement an agricultural inputs support scheme. Christian Care said it is now providing farmers in the Midlands with fertilizer and seeds. Experts said most farmers are struggling to obtain inputs because the government has cut back on programs to finance planting, instead urging farmers to borrow from banks. But banks insist on land as collateral, though all farmland has been nationalized. Even farmers resettled under land reform since 2000 have only so-called offer letters granting them working rights, but banks will accept neither these nor livestock as collateral. Christian Care Director Forbes Matonga reported that his is one of seven organizations reaching out to small farmers under the FAO program. As the rainy season begins, cholera is also now a growing concern. Residents of Zimbabwe, fearing a repeat of the 2008 epidemic that killed thousands, are struggling to complete precautionary measures. 33 | -------------NEW ARTICLE----------------- 34 | After 100 days in office, United States president Barack Obama gave a speech on Wednesday, speaking about the swine influenza outbreak and the struggling economy, both described by the Los Angles Times as "two wars." He used a prime time television slot to showcase his message throughout the United States. During his speech, he said, "If you could tell me right now when I walked into this office... that all you had to worry about was Iraq, Afghanistan, North Korea, getting healthcare passed, figuring out how to deal with energy independence, deal with Iran and a pandemic flu, I would take that deal. I would love a nice, lean portfolio to deal with, but that's not the hand that's been dealt us." Obama also said the economy was not the only problem. There are threats to the country including "...terrorism to nuclear proliferation to pandemic flu." Regarding the swine influenza outbreak, he said that the U.S. / Mexico border will not be closed because closing the border does not fix any problems, claiming that this method did not work in the past. Instead, he said that the best method for preventing the spread of the flu is hand washing, covering one's mouth while coughing, and staying home when one feels sick. The Los Angeles Times described Obama "more like school nurse in chief than commander in chief." On the topic of waterboarding, Obama said, "I do believe that it is torture." 35 | -------------NEW ARTICLE----------------- 36 | The World Organization for Animal Health (OIE) and the Food and Agriculture Organization, both part of the United Nations, have stated that some countries, particularly China, Indonesia and some countries in Africa are "under-reporting" the number of human cases of the deadly H5N1 Avian Flu (Bird Flu) virus, but also said that the countries are not hiding them "deliberately." "We know that some countries might be under-reporting ... most do not do it deliberately. We are concerned about China and Indonesia and Africa because the virus seems to be so widespread that we could not get all the information. It is difficult to know about each individual outbreak in a back yard," said Doctor Christianne Bruschke, in Rome, Italy on Wednesday and who is head of the OIE's Bird Flu taskforce. Bruschke also said that farmers lack the education they need on the virus and need to be reimbursed financially for any education they need citing that the "richer nations" should help fund the education they need and that lack of veterinary clinics, distance to them, and time are also to blame for the under-reporting. In Africa, "farmers will probably not report sick animals," said Bruschke. "Their veterinary services are very weak and many countries do not have laboratory facilities - we have all the ingredients there that could lead to under-reporting," she added. Bruschke also said that Indonesia may not be reporting all human or animal cases also stating that the virus is "permanently infecting poultry" in the country which makes it increasingly difficult for anyone to report outbreaks. "I think it could be the case because in certain regions the virus is getting more or less endemic, so in regions like Java, they might not report every single outbreak anymore," said Bruschke. According to Bruschke, China is cooperating but she also said that "China is a very big country" and that there are cases of infections in wild birds. "We sometimes see the outbreaks in wild animals - they will not always detect them. There is also not a very good compensation scheme in place so we feel there might be under-reporting," said Bruschke. 37 | -------------NEW ARTICLE----------------- 38 | The H1N1 outbreak of swine flu, which began in Mexico this April, has now spread across the globe. There have been at least 3,330 deaths from the swine flu since the virus started spreading, out of almost 316,000 total reported cases. Nine countries — Australia, Brazil, France, Italy, New Zealand, Norway, Switzerland, the UK, and the US — have promised to send ten percent of their antiviral vaccine supply to other countries, should the latter be in need of it. The plan was agreed to "in recognition that diseases know no borders and that the health of the American people is inseparable from the health of people around the world," a statement by the US government read. In an in-depth report, Wikinews takes a look at how the disease has affected countries around the world. As of Wednesday, Brazil has registered 899 deaths from the swine flu, making it the hardest-hit country in terms of fatalities. The city of Sao Paulo reported 327 deaths, and Rio de Janeiro 84. However, the country's health ministry also said that the rate of serious cases "fell for the fifth straight week." Brazil had surpassed the United States, which has 593 deaths, in number of total fatalities from the outbreak late in August. Argentina, Brazil's neighbour to the south, has 512 deaths from the H1N1 virus. 10,000 cases of swine flu were confirmed across China since the outbreak began. The number of infections seems to be increasing quickly. Communications director for the World Health Organisation Vivian Tan said that "in the last week or so, the increase has been quite quick," attributing the rise to a small decrease in temperatures as fall sets in, as well as students returning to school after summer breaks. China's official news agency Xinhua reported that 1,118 new cases of the influenza were reported in a two-day period earlier in the week, adding that a vast majority of the cases had been transmitted in China, not by persons entering the country from abroad. All 31 of China's provinces have reported instances of the flu. The disease initially seemed to be limited to large cities, but recently has started moving into more rural areas. No casualties from the swine flu have yet been confirmed in China. The office of France's president Nicolas Sarkozy said that the country would pledge up to one tenth or nine million of its 94 million antiviral vaccine doses to the World Health Organisation, to be distributed to countries with fewer vaccine supplies if needed. International solidarity "will be a determining factor in reducing the health, economic and social impact of the pandemic," according to a statement released by the government. On Wednesday, eleven people had been reported dead from the virus in India, taking the country's death toll up to 212 people. The number of people infected with the influenza is now estimated at 6,800. India's health ministry on Tuesday said that the Tamiflu drug would be on sale in the open market within seven days, allowing for a "restricted sale" of the drug. An unnamed official said that "it is expected that within the next five to seven days, both the drugs would be available in the retail market through identified chemists against proper medical prescriptions. "Taking into account the current spread of the influenza A(H1N1) in the country, the health ministry has decided that retail sale of Tamiflu and Zanamivir should be allowed in the country but in a regulated manner," he said. Previously, distribution of Tamiflu was prohibited by the government, and access to it was only available through public health institutions. At least 70 people in Kenya have the swine flu, according to local health official. In the latest outbreak, twenty high school students came down with the virus and had to be quarantined on Thursday. "A majority of the affected students who are in Forms One and Two were treated and advised to remain under bed rest to minimise further spread of the disease among the student community," said the director of Public Health, Dr. Shanaaz Shariff. However, he said that the students' illness was "not too serious to warrant hospitalisation." Security guards were placed around the school the students were isolated in, with orders only to allow medical personnel to enter the premises. Kenya's capital Nairobi has been the worst hit by the flu, having reported forty cases. Other cities affected by the flu are Kisumu and Rift Valley, who have reported eighteen and ten cases of the H1N1 virus, respectively. Mexico, the country in which the outbreak initially started, has 25,214 reported cases and 217 fatalities from the virus. Some recent cases have forced schools to close down. Jose Angel Cordova, the Mexican health secretary, said that the virus could infect as many as five million of Mexico's 107 million people, and, in a worst-case scenario, cause up to 2,000 deaths. His estimate is higher than his previous prediction of 1 million cases and 1,000 deaths, made last month. About five thousand new cases of swine flu were reported in the United Kingdom in recent weeks, reversing a declining trend in the number of new infections. Health officials have suggested this could lead up to a second outbreak of the virus. "We don't know whether this is the start of the next big wave that we were expecting this autumn but it is certainly something that's giving us concern. It will probably be a week or two before we see whether this increase is sustained." said Liam Donaldson, the Chief Medical Officer. Health authorities have said that at least 25 cases appear to have been resistant to the Tamiflu drug prescribed to treat the illness. Donaldson said that "the positive side of it is that so far these have not been strains that have then gone on and affected other patients, they have stayed with the patient in which they were isolated. What would worry us is if we got a resistant strain that then started infecting people like the rest of the cases of flu that have occurred." The UK is one of several countries that have pledged up to one tenth of their vaccine stock to to other countries if they are in need of more supplies. "Britain recognizes that H1N1 is a global pandemic which requires a global response," the International Development Secretary, Douglas Alexander, said. "Solidarity with other nations is vital, particularly the poorest who may be most vulnerable and have least capacity to respond." The US government recently bought 195 million doses of swine flu vaccine. Health Care Secretary Kathleen Sebelius said that free shots will be given out early in October. The vaccination is to be voluntary, but priority will be given to certain groups, such as toddlers and children, adults over the age of 65, and pregnant women, who are considered especially vulnerable to the virus. "We remain confident that the United States will have sufficient doses of the vaccine to ensure that every American who wants a vaccine is able to receive one," a White House statement said. As of September 16, the US had 593 deaths from the flu. 144 people in Vietnam were diagnosed with the swine flu on Wednesday, bringing the total number of infected people in the country to 5,648. This week, the number of affected people has increased by 1104 infections or 6.3%. Nguyen Tran Hien, the director of the Central Institute of Hygiene Epidemiology, predicted that the swine flu would peak at the end of 2009 and the beginning 2010. The Vietnamese Ministry of Health called for more research into a swine flu vaccine, and urged the the National Steering Board on Flu Prevention in Humans to give out more doses of the drug Tamiflu to areas hardest hit by the flu. 39 | -------------NEW ARTICLE----------------- 40 | On Saturday local villagers claimed a meteorite slammed into a field outside of Carancas, near Lake Titicaca in the Puno region of Peru on the border of Bolivia. It emitted a sweet but noxious odor. It has now been blamed for a mass illness affecting roughly 200 villagers with "nausea, vomiting, digestive problems and general sickness," according to a local health department official, Jorge López. "Boiling water started coming out of the crater and particles of rock and cinders were found nearby. Residents are very concerned," said López. Police officers who went to investigate the meteorite are among those who have fallen ill and been taken to Desaguadero hospital. The impact of the meteorite left a crater 18 feet deep and 30 yards across in the Andean territory that is home to less than 1,000 people. Originally, the villagers thought a plane had crashed. Under consideration is the declaration of a state of emergency. Peruvian Nuclear Energy Institute engineer Renan Ramirez said scientists who went to investigate the crater found no indication of radiation, although, the fumes from the crater were so pungent that one scientist said his throat and nose were irritated despite his use of a mask. Villagers are said to be avoiding the local water out of fear of contamination. Sulfur, arsenic and other elements common in meteorites can react with ground water to produce fumes. Ursula Marvin, a meteor expert at the Smithsonian Astrophysical Observatory in Massachusetts, said a meteorite "wouldn't get much gas out of the earth" and that a more likely explanation for the health problems was the dust cloud caused when the rock hit the Earth. Other explanations abound. Don Yeomans, head of the Near Earth Object Program at NASA's Jet Propulsion Laboratory, said "Statistically, it's far more likely to have come from below than from above," noting that meteorites do not give off smells and it is likely the result of hydrothermal activity, such as a local gas explosion. Dr Caroline Smith, a meteorite expert with the Natural History Museum in London, thought more likely the villagers saw a common fireball and in the process of investigating it, did not find a 'crater' but "a lake of sedimentary deposit, which may be full of smelly, methane rich organic matter." And one science blogger, David Syzdek, theorized that in fact the crater is a mud volcano "that is producing toxic gasses [sic] such as methane, water, and perhaps other hydrocarbons. Many hydrocarbons can cause illness." 41 | -------------NEW ARTICLE----------------- 42 | At least two children in Baghdad, Iraq have died after eating cake poisoned with thallium and at least nine others remain ill. The cake was given to people at an Iraqi sports club near the capital. The Secretary of the Iraqi Air Force and his daughter are also among the victims. All are currently receiving treatment in Amman, Jordan as the antidote for the deadly poison is not available in Iraqi hospitals. The United Kingdom has flown in the antidotes and treatments for the victims, several of which are seriously ill. "This is a disturbing incident," said a police spokesman. Thallium is used in poisons to kill rats and insects and is also considered the assassin's drug or "The Poisoner's Poison" because it is tasteless. Former Iraqi dictator Saddam Hussein used the poison against many of his enemies, and it has not been used since his rule of the country. So far, it is not known how the cakes were poisoned, but the Air force Secretary says that the cakes were seemingly delivered to the club as a "goodwill gesture." Two officials at the club then took the cake home, where members of the families ate it. He now believes someone conspired to kill him and his family. "The use of thallium in this way appears to show that someone in Adhamiya is reviving the techniques of the mukhabarat [Saddam's secret police]. What happens if al-Qaida gets the know-how? We are urgently trying to discover how much thallium is out there and who would know how to utilize it," added the spokesman. An investigation is ongoing, but police say that the cakes were made at a local bakery in the Adhamiyah district of Baghdad. Police also say that the bakery has recently been a host for Sunni militants. -------------------------------------------------------------------------------- /data/words2index.pkl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/milangritta/Geocoding-with-Map-Vector/69e6e590c56930ed80d346f6c2da6d58056182e3/data/words2index.pkl -------------------------------------------------------------------------------- /geoparse.py: -------------------------------------------------------------------------------- 1 | import cPickle 2 | import codecs 3 | import sqlite3 4 | from genericpath import isfile 5 | from os import listdir 6 | import spacy 7 | import numpy as np 8 | from geopy.distance import great_circle 9 | from keras.models import load_model 10 | from preprocessing import index_to_coord, ENCODING_MAP_1x1, OUTLIERS_MAP_1x1, get_coordinates 11 | from preprocessing import CONTEXT_LENGTH, pad_list, TARGET_LENGTH, UNKNOWN, REVERSE_MAP_2x2 12 | from text2mapVec import text2mapvec 13 | 14 | model = load_model("../data/weights") # weights to be downloaded from Cambridge Uni repo, see GitHub. 15 | nlp = spacy.load(u'en_core_web_lg') # or spacy.load(u'en') depending on your Spacy Download (simple or full) 16 | conn = sqlite3.connect(u'../data/geonames.db').cursor() # this DB can be downloaded using the GitHub link 17 | padding = nlp(u"0")[0] # Do I need to explain? :-) 18 | word_to_index = cPickle.load(open(u"data/words2index.pkl")) # This is the vocabulary file 19 | 20 | for word in nlp.Defaults.stop_words: # This is only necessary if you use the full Spacy English model 21 | lex = nlp.vocab[word] # so if you use spacy.load(u'en'), you can comment this out. 22 | lex.is_stop = True 23 | 24 | 25 | def geoparse(text): 26 | """ 27 | This function allows one to geoparse text i.e. extract toponyms (place names) and disambiguate to coordinates. 28 | :param text: to be parsed 29 | :return: currently only prints results to the screen, feel free to modify to your task 30 | """ 31 | doc = nlp(text) # NER with Spacy NER 32 | for entity in doc.ents: 33 | if entity.label_ in [u"GPE", u"FACILITY", u"LOC", u"FAC", u"LOCATION"]: 34 | name = entity.text if not entity.text.startswith('the') else entity.text[4:].strip() 35 | start = entity.start_char if not entity.text.startswith('the') else entity.start_char + 4 36 | end = entity.end_char 37 | near_inp = pad_list(CONTEXT_LENGTH / 2, [x for x in doc[max(0, entity.start - CONTEXT_LENGTH / 2):entity.start]], True, padding) + \ 38 | pad_list(CONTEXT_LENGTH / 2, [x for x in doc[entity.end: entity.end + CONTEXT_LENGTH / 2]], False, padding) 39 | far_inp = pad_list(CONTEXT_LENGTH / 2, [x for x in doc[max(0, entity.start - CONTEXT_LENGTH):max(0, entity.start - CONTEXT_LENGTH / 2)]], True, padding) + \ 40 | pad_list(CONTEXT_LENGTH / 2, [x for x in doc[entity.end + CONTEXT_LENGTH / 2: entity.end + CONTEXT_LENGTH]], False, padding) 41 | map_vector = text2mapvec(doc=near_inp + far_inp, mapping=ENCODING_MAP_1x1, outliers=OUTLIERS_MAP_1x1, polygon_size=1, db=conn, exclude=name) 42 | 43 | context_words, entities_strings = [], [] 44 | target_string = pad_list(TARGET_LENGTH, [x.text.lower() for x in entity], True, u'0') 45 | target_string = [word_to_index[x] if x in word_to_index else word_to_index[UNKNOWN] for x in target_string] 46 | for words in [near_inp, far_inp]: 47 | for word in words: 48 | if word.text.lower() in word_to_index: 49 | vec = word_to_index[word.text.lower()] 50 | else: 51 | vec = word_to_index[UNKNOWN] 52 | if word.ent_type_ in [u"GPE", u"FACILITY", u"LOC", u"FAC", u"LOCATION"]: 53 | entities_strings.append(vec) 54 | context_words.append(word_to_index[u'0']) 55 | elif word.is_alpha and not word.is_stop: 56 | context_words.append(vec) 57 | entities_strings.append(word_to_index[u'0']) 58 | else: 59 | context_words.append(word_to_index[u'0']) 60 | entities_strings.append(word_to_index[u'0']) 61 | 62 | prediction = model.predict([np.array([context_words]), np.array([context_words]), np.array([entities_strings]), 63 | np.array([entities_strings]), np.array([map_vector]), np.array([target_string])]) 64 | prediction = index_to_coord(REVERSE_MAP_2x2[np.argmax(prediction[0])], 2) 65 | candidates = get_coordinates(conn, name) 66 | 67 | if len(candidates) == 0: 68 | print(u"Don't have an entry for", name, u"in GeoNames") 69 | continue 70 | 71 | max_pop = candidates[0][2] 72 | best_candidate = [] 73 | bias = 0.905 # Tweak the parameter depending on the domain you're working with. 74 | # Less than 0.9 suitable for ambiguous text, more than 0.9 suitable for less ambiguous locations, see paper 75 | for candidate in candidates: 76 | err = great_circle(prediction, (float(candidate[0]), float(candidate[1]))).km 77 | best_candidate.append((err - (err * max(1, candidate[2]) / max(1, max_pop)) * bias, (float(candidate[0]), float(candidate[1])))) 78 | best_candidate = sorted(best_candidate, key=lambda (a, b): a)[0] 79 | 80 | # England,, England,, 51.5,, -0.11,, 669,, 676 || - use evaluation script to test correctness 81 | print name, start, end 82 | print u"Coordinates:", best_candidate[1] 83 | 84 | 85 | # Example usage of the geoparse function below reading from a directory and parsing all files. 86 | directory = u"/Users/milangritta/PycharmProjects/data/lgl/" 87 | files = [f for f in listdir(directory) if isfile(directory + f)] 88 | for f in files: 89 | for line in codecs.open(directory + f, encoding="utf-8"): 90 | print line 91 | geoparse(line) 92 | -------------------------------------------------------------------------------- /geovirus.py: -------------------------------------------------------------------------------- 1 | import codecs 2 | import random 3 | import sqlite3 4 | import xml.etree.ElementTree as ET 5 | from collections import Counter 6 | import numpy 7 | from geopy.distance import great_circle 8 | from preprocessing import get_coordinates 9 | 10 | # -------------------------------------------- ERROR CHECKING ---------------------------------------------- 11 | 12 | if False: 13 | """ 14 | Check for XML formatting, duplicate articles, URLs, coordinate distances to Geonames database, 15 | correct indexing of location names i.e. start and end character positions. 16 | """ 17 | tree = ET.parse(u'data/GeoVirus.xml') 18 | conn = sqlite3.connect(u'../data/geonames.db') 19 | c = conn.cursor() 20 | root = tree.getroot() 21 | duplicates = set() 22 | for article in root: 23 | text = article.find('text').text 24 | if text in duplicates: 25 | raise Exception(u'Duplicate titles/sources!') 26 | else: 27 | duplicates.add(text) 28 | for location in article.find('locations'): 29 | start = location.find('start').text 30 | end = location.find('end').text 31 | name = location.find('name').text 32 | url = location.find('page') 33 | if url.text != u"N/A": 34 | if url is None or not url.text.startswith(u"https://en.wikipedia.org/wiki/"): 35 | raise Exception(u"URL corrupted!") 36 | chunk = text[int(start) - 1: int(end) - 1] 37 | if chunk != name: 38 | raise Exception(chunk + ", " + name) 39 | if location.find('altName') is not None: 40 | name = location.find('altName').text 41 | lat = location.find('lat').text 42 | lon = location.find('lon').text 43 | coords = get_coordinates(c, name) 44 | dist = 10000.0 45 | for coord in coords: 46 | gap = great_circle((float(lat), float(lon)), (coord[0], coord[1])).km 47 | if gap < dist: 48 | dist = gap 49 | if dist > 1000: 50 | print u"Distance is large, please check if this is normal.", name, url, dist, lat, lon 51 | 52 | # -------------------------------------------------- NUMBERS ------------------------------------------------------- 53 | 54 | if False: 55 | """ 56 | Generate essential stats describing the nature of the dataset. Reported in the publication. 57 | """ 58 | tree = ET.parse('data/GeoVirus.xml') 59 | root = tree.getroot() 60 | counter, continents, words, articles = [], 0, [], 0 61 | for article in root: 62 | articles += 1 63 | text = article.find("text").text 64 | words.extend(text.split()) 65 | for location in article.findall("locations/location"): 66 | name = location.find("name") 67 | cont = location.find("continent") 68 | if cont is not None: 69 | continents += 1 70 | counter.append(name.text) 71 | print "Total Locations:", len(counter) 72 | counter = Counter(counter) 73 | print "Unique Locations:", len(counter) 74 | print "Most Common:", counter.most_common() 75 | print "Continents", continents 76 | counter = [j for (i, j) in counter.most_common()] 77 | print "Mean:", numpy.mean(counter) 78 | print "Median:", numpy.median(counter) 79 | print "Articles:", articles 80 | print "Total words:", len(words) 81 | 82 | 83 | # ---------------------------------------------- GENERATION ------------------------------------------------ 84 | 85 | if False: 86 | """ 87 | Before running the function, please create a directory called "geovirus" outside of the loc2vec directory. 88 | This function is used to convert the XML file into (1.) a directory of files where each file contains the 89 | text of each article i.e. 229 files will be created. (2.) a file "geovirus_gold.txt" containing the gold answers 90 | for each article. These two outputs will be used to generate evaluation files in preprocessing.py 91 | """ 92 | tree = ET.parse(u"data/GeoVirus.xml") 93 | root = tree.getroot() 94 | f = codecs.open(u"data/geovirus_gold.txt", "w", "utf-8") 95 | c = 0 96 | counter = [] 97 | for child in root: 98 | text = child.find('text').text 99 | gold_tops = [] 100 | for location in child.findall('./locations/location'): 101 | start = location.find("start") 102 | end = location.find("end") 103 | name = location.find("name") 104 | if location.find('altName') is not None: 105 | alt_name = location.find('altName') 106 | else: 107 | alt_name = name 108 | counter.append(name.text) 109 | lat = location.find("lat") 110 | lon = location.find("lon") 111 | gold_tops.append(alt_name.text + ",," + name.text + ",," + lat.text + ",," + lon.text + ",," + start.text + ",," + end.text) 112 | for t in gold_tops: 113 | f.write(t + "||") 114 | f.write("\n") 115 | f_out = codecs.open(u"../data/geovirus/" + str(c), 'w', "utf-8") # Files saved by numbers 116 | f_out.write(text) 117 | f_out.close() 118 | c += 1 119 | f.close() 120 | counter = Counter(counter) 121 | print counter.most_common() 122 | 123 | # --------------------------------------SUBSAMPLING FOR INTER-ANNOTATOR AGREEMENT-------------------------------------- 124 | 125 | if False: 126 | """ 127 | Generate a 10% random sample for the Inter Annotator Agreement figures. 128 | """ 129 | iaa_check = codecs.open(u"data/iaa_check.txt", "w", "utf-8") 130 | iaa_test = codecs.open(u"data/iaa_test.txt", "w", "utf-8") 131 | tree = ET.parse(u'data/GeoVirus.xml') 132 | root = tree.getroot() 133 | 134 | for article in root: 135 | if random.randint(1, 10) > 9: 136 | text = article.find("text").text 137 | iaa_test.write("-------------NEW ARTICLE-----------------\n") 138 | iaa_test.write(text + "\n") 139 | print_count = 0 140 | for loc in article.findall("./locations/location"): 141 | print_count += 1 142 | start = int(loc.find("start").text) 143 | iaa_check.write(loc.find("page").text + "\n") 144 | iaa_check.write(loc.find("start").text + "\n") 145 | iaa_check.write(loc.find("name").text + "\n") 146 | if print_count <= 3: 147 | iaa_test.write("-----------\n") 148 | iaa_test.write("LOCATION NAME -> Asia\n") 149 | iaa_test.write("LINK -> https://en.wikipedia.org/wiki/Asia\n") 150 | iaa_test.write("START CHARACTER -> 100\n") 151 | 152 | # -----------------------------------------COMPUTING INTER-ANNOTATOR AGREEMENT--------------------------------------- 153 | 154 | if False: 155 | """ 156 | Compute IAA, print out overlaps and disagreements, then calculate IAA figures manually. 157 | """ 158 | iaa_answers = [] 159 | for index, line in enumerate(codecs.open(u"data/iaa_answers.txt", "r", "utf-8"), start=1): 160 | if index % 3 == 0: 161 | iaa_answers.append((url, start, line.strip())) 162 | elif index % 3 == 1: 163 | url = line.strip() 164 | else: 165 | start = int(line) + 1 166 | 167 | iaa_check = [] 168 | for index, line in enumerate(codecs.open(u"data/iaa_check.txt", "r", "utf-8"), start=1): 169 | if index % 3 == 0: 170 | iaa_check.append((url, start, line.strip())) 171 | elif index % 3 == 1: 172 | url = line.strip() 173 | else: 174 | start = int(line) 175 | 176 | intersection = Counter(iaa_check) & Counter(iaa_answers) 177 | print len(intersection) 178 | check = Counter(iaa_check) - intersection 179 | answers = Counter(iaa_answers) - intersection 180 | iaa_check = list(check.elements()) 181 | iaa_answers = list(answers.elements()) 182 | print iaa_check 183 | print iaa_answers 184 | 185 | # ----------------------------------------- END ------------------------------------------- 186 | -------------------------------------------------------------------------------- /melbourne-augmenting-geocoding.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/milangritta/Geocoding-with-Map-Vector/69e6e590c56930ed80d346f6c2da6d58056182e3/melbourne-augmenting-geocoding.pdf -------------------------------------------------------------------------------- /preprocessing.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | import codecs 3 | import cPickle 4 | from collections import Counter 5 | import matplotlib.pyplot as plt 6 | import spacy 7 | import numpy as np 8 | import sqlite3 9 | from geopy.distance import great_circle 10 | from matplotlib import pyplot, colors 11 | 12 | 13 | # -------- GLOBAL CONSTANTS AND VARIABLES -------- # 14 | BATCH_SIZE = 64 15 | CONTEXT_LENGTH = 200 # each side of target entity 16 | UNKNOWN = u"" 17 | EMBEDDING_DIMENSION = 50 18 | TARGET_LENGTH = 15 19 | ENCODING_MAP_1x1 = cPickle.load(open(u"data/1x1_encode_map.pkl")) # We need these maps 20 | ENCODING_MAP_2x2 = cPickle.load(open(u"data/2x2_encode_map.pkl")) # and the reverse ones 21 | REVERSE_MAP_1x1 = cPickle.load(open(u"data/1x1_reverse_map.pkl")) # to handle the used and 22 | REVERSE_MAP_2x2 = cPickle.load(open(u"data/2x2_reverse_map.pkl")) # unused map_vector polygons. 23 | OUTLIERS_MAP_1x1 = cPickle.load(open(u"data/1x1_outliers_map.pkl")) # Outliers are redundant polygons that 24 | OUTLIERS_MAP_2x2 = cPickle.load(open(u"data/2x2_outliers_map.pkl")) # have been removed but must also be handled. 25 | # -------- GLOBAL CONSTANTS AND VARIABLES -------- # 26 | 27 | 28 | def print_stats(accuracy): 29 | """ 30 | Prints Mean, Median, AUC and acc@161km for the list. 31 | :param accuracy: a list of geocoding errors 32 | """ 33 | print("==============================================================================================") 34 | print(u"Median error:", np.median(sorted(accuracy))) 35 | print(u"Mean error:", np.mean(accuracy)) 36 | accuracy = np.log(np.array(accuracy) + 1) 37 | k = np.log(161) 38 | print u"Accuracy to 161 km: ", sum([1.0 for dist in accuracy if dist < k]) / len(accuracy) 39 | print u"AUC = ", np.trapz(accuracy) / (np.log(20039) * (len(accuracy) - 1)) # Trapezoidal rule. 40 | print("==============================================================================================") 41 | 42 | 43 | def pad_list(size, a_list, from_left, padding): 44 | """ 45 | Utility function that pads a list with any given padding. 46 | :param size: the final length of the list i.e. pad up to size 47 | :param a_list: the list to pad 48 | :param from_left: True to pad from the left, False to pad from the right 49 | :param padding: whatever you want to use for padding, example "0" 50 | :return: the padded list 51 | """ 52 | while len(a_list) < size: 53 | if from_left: 54 | a_list = [padding] + a_list 55 | else: 56 | a_list += [padding] 57 | return a_list 58 | 59 | 60 | def coord_to_index(coordinates, polygon_size): 61 | """ 62 | Convert coordinates into an array (world representation) index. Use that to modify map_vector polygon value. 63 | :param coordinates: (latitude, longitude) to convert to the map vector index 64 | :param polygon_size: integer size of the polygon? i.e. the resolution of the world 65 | :return: index pointing into map_vector array 66 | """ 67 | latitude = float(coordinates[0]) - 90 if float(coordinates[0]) != -90 else -179.99 # The two edge cases must 68 | longitude = float(coordinates[1]) + 180 if float(coordinates[1]) != 180 else 359.99 # get handled differently! 69 | if longitude < 0: 70 | longitude = -longitude 71 | if latitude < 0: 72 | latitude = -latitude 73 | x = int(360 / polygon_size) * int(latitude / polygon_size) 74 | y = int(longitude / polygon_size) 75 | return x + y if 0 <= x + y <= int(360 / polygon_size) * int(180 / polygon_size) else Exception(u"Shock horror!!") 76 | 77 | 78 | def index_to_coord(index, polygon_size): 79 | """ 80 | Convert index (output of the prediction model) back to coordinates. 81 | :param index: of the polygon/tile in map_vector array (given by model prediction) 82 | :param polygon_size: size of each polygon/tile i.e. resolution of the world 83 | :return: pair of (latitude, longitude) 84 | """ 85 | x = int(index / (360 / polygon_size)) 86 | y = index % int(360 / polygon_size) 87 | if x > int(90 / polygon_size): 88 | x = -int((x - (90 / polygon_size)) * polygon_size) 89 | else: 90 | x = int(((90 / polygon_size) - x) * polygon_size) 91 | if y < int(180 / polygon_size): 92 | y = -int(((180 / polygon_size) - y) * polygon_size) 93 | else: 94 | y = int((y - (180 / polygon_size)) * polygon_size) 95 | return x, y 96 | 97 | 98 | def get_coordinates(con, loc_name): 99 | """ 100 | Access the database to retrieve coordinates and other data from DB. 101 | :param con: sqlite3 database cursor i.e. DB connection 102 | :param loc_name: name of the place 103 | :return: a list of tuples [(latitude, longitude, population, feature_code), ...] 104 | """ 105 | result = con.execute(u"SELECT METADATA FROM GEO WHERE NAME = ?", (loc_name.lower(),)).fetchone() 106 | if result: 107 | result = eval(result[0]) # Do not remove the sorting, the function below assumes sorted results! 108 | return sorted(result, key=lambda (a, b, c, d): c, reverse=True) 109 | else: 110 | return [] 111 | 112 | 113 | def construct_map_vector(a_list, polygon_size, mapping, outliers): 114 | """ 115 | Build the map_vector vector representation from a_list of location data. 116 | :param a_list: of tuples [(latitude, longitude, population, feature_code), ...] 117 | :param polygon_size: what's the resolution? size of each polygon in degrees. 118 | :param mapping: one of the transformation maps 1x1 or 2x2 119 | :param outliers: the outlier map, 1x1 or 2x2 (must match resolution or mapping above) 120 | :return: map_vector representation 121 | """ 122 | map_vector = np.zeros(len(mapping), ) 123 | if len(a_list) == 0: 124 | return map_vector 125 | max_pop = a_list[0][2] if a_list[0][2] > 0 else 1 126 | for s in a_list: 127 | index = coord_to_index((s[0], s[1]), polygon_size) 128 | if index in mapping: 129 | index = mapping[index] 130 | else: 131 | index = mapping[outliers[index]] 132 | map_vector[index] += float(max(s[2], 1)) / max_pop 133 | return map_vector / map_vector.max() if map_vector.max() > 0.0 else map_vector 134 | 135 | 136 | def construct_map_vector_full_scale(a_list, polygon_size): 137 | """ 138 | This function is similar to the above BUT it builds map_vector WITHOUT removing redundant polygons. 139 | :param a_list: of tuples [(latitude, longitude, population, feature_code), ...] 140 | :param polygon_size: size of each polygon in degrees i.e 1x1 or 2x2 141 | :return: map_vector (full scale) i.e. without removing redundant polygons, used for visualisation in 2D 142 | """ 143 | map_vector = np.zeros(int(360 / polygon_size) * int(180 / polygon_size)) 144 | if len(a_list) == 0: 145 | return map_vector 146 | max_pop = a_list[0][2] if a_list[0][2] > 0 else 1 147 | for s in a_list: 148 | index = coord_to_index((s[0], s[1]), polygon_size) 149 | map_vector[index] += float(max(s[2], 1)) / max_pop 150 | return map_vector / map_vector.max() if map_vector.max() > 0.0 else map_vector 151 | 152 | 153 | def merge_lists(lists): 154 | """ 155 | Utility function to merge multiple lists. 156 | :param lists: a list of lists to be merged 157 | :return: one single list with all items from above list of lists 158 | """ 159 | out = [] 160 | for l in lists: 161 | out.extend(l) 162 | return out 163 | 164 | 165 | def populate_sql(): 166 | """ 167 | Create and populate the sqlite3 database with GeoNames data. Requires Geonames dump. 168 | No need to run this function, I share the database as a separate dump on GitHub (see link). 169 | """ 170 | geo_names = {} 171 | p_map = {"PPLC": 100000, "PCLI": 100000, "PCL": 100000, "PCLS": 10000, "PCLF": 10000, "CONT": 100000, "RGN": 100000} 172 | 173 | for line in codecs.open(u"../data/allCountries.txt", u"r", encoding=u"utf-8"): 174 | line = line.split("\t") 175 | feat_code = line[7] 176 | class_code = line[6] 177 | pop = int(line[14]) 178 | for name in [line[1], line[2]] + line[3].split(","): 179 | name = name.lower() 180 | if len(name) != 0: 181 | if name in geo_names: 182 | already_have_entry = False 183 | for item in geo_names[name]: 184 | if great_circle((float(line[4]), float(line[5])), (item[0], item[1])).km < 100: 185 | if item[2] >= pop: 186 | already_have_entry = True 187 | if not already_have_entry: 188 | pop = get_population(class_code, feat_code, p_map, pop) 189 | geo_names[name].add((float(line[4]), float(line[5]), pop, feat_code)) 190 | else: 191 | pop = get_population(class_code, feat_code, p_map, pop) 192 | geo_names[name] = {(float(line[4]), float(line[5]), pop, feat_code)} 193 | 194 | conn = sqlite3.connect(u'../data/geonames.db') 195 | c = conn.cursor() 196 | # c.execute("CREATE TABLE GEO (NAME VARCHAR(100) PRIMARY KEY NOT NULL, METADATA VARCHAR(5000) NOT NULL);") 197 | c.execute(u"DELETE FROM GEO") # alternatively, delete the database file. 198 | conn.commit() 199 | 200 | for gn in geo_names: 201 | c.execute(u"INSERT INTO GEO VALUES (?, ?)", (gn, str(list(geo_names[gn])))) 202 | print(u"Entries saved:", len(geo_names)) 203 | conn.commit() 204 | conn.close() 205 | 206 | 207 | def get_population(class_code, feat_code, p_map, pop): 208 | """ 209 | Utility function to eliminate code duplication. Nothing of much interest, methinks. 210 | :param class_code: Geonames code for the class of location 211 | :param feat_code: Geonames code for the feature type of an database entry 212 | :param p_map: dictionary mapping feature codes to estimated population 213 | :param pop: population count 214 | :return: population (modified if class code is one of A, P or L. 215 | """ 216 | if pop == 0 and class_code in ["A", "P", "L"]: 217 | pop = p_map.get(feat_code, 0) 218 | return pop 219 | 220 | 221 | def generate_training_data(): 222 | """ 223 | Prepare Wikipedia training data. Please download the required files from GitHub. 224 | Files: geonames.db and geowiki.txt both inside the data folder (see README) 225 | Alternatively, create your own with http://medialab.di.unipi.it/wiki/Wikipedia_Extractor 226 | """ 227 | conn = sqlite3.connect(u'../data/geonames.db') 228 | c = conn.cursor() 229 | nlp = spacy.load(u'en') # or spacy.load(u'en_core_web_lg') depending on your Spacy Download (simple, full) 230 | padding = nlp(u"0")[0] 231 | inp = codecs.open(u"../data/geowiki.txt", u"r", encoding=u"utf-8") 232 | o = codecs.open(u"../data/train_wiki.txt", u"w", encoding=u"utf-8") 233 | lat, lon = u"", u"" 234 | target, string = u"", u"" 235 | skipped = 0 236 | 237 | for line in inp: 238 | if len(line.strip()) == 0: 239 | continue 240 | limit = 0 241 | if line.startswith(u"NEW ARTICLE::"): 242 | if len(string.strip()) > 0 and len(target) != 0: 243 | locations_near, locations_far = [], [] 244 | doc = nlp(string) 245 | for d in doc: 246 | if d.text == target[0]: 247 | if u" ".join(target) == u" ".join([t.text for t in doc[d.i:d.i + len(target)]]): 248 | near_inp = pad_list(CONTEXT_LENGTH / 2, [x for x in doc[max(0, d.i - CONTEXT_LENGTH / 2):d.i]], True, padding) \ 249 | + pad_list(CONTEXT_LENGTH / 2, [x for x in doc[d.i + len(target): d.i + len(target) + CONTEXT_LENGTH / 2]], False, padding) 250 | far_inp = pad_list(CONTEXT_LENGTH / 2, [x for x in doc[max(0, d.i - CONTEXT_LENGTH):max(0, d.i - CONTEXT_LENGTH / 2)]], True, padding) \ 251 | + pad_list(CONTEXT_LENGTH / 2, [x for x in doc[d.i + len(target) + CONTEXT_LENGTH / 2: d.i + len(target) + CONTEXT_LENGTH]], False, padding) 252 | near_out, far_out = [], [] 253 | location = u"" 254 | for (out_list, in_list, is_near) in [(near_out, near_inp, True), (far_out, far_inp, False)]: 255 | for index, item in enumerate(in_list): 256 | if item.ent_type_ in [u"GPE", u"FACILITY", u"LOC", u"FAC", u"LOCATION"]: 257 | if item.ent_iob_ == u"B" and item.text.lower() == u"the": 258 | out_list.append(item.text.lower()) 259 | else: 260 | location += item.text + u" " 261 | out_list.append(u"**LOC**" + item.text.lower()) 262 | elif item.ent_type_ in [u"PERSON", u"DATE", u"TIME", u"PERCENT", u"MONEY", 263 | u"QUANTITY", u"CARDINAL", u"ORDINAL"]: 264 | out_list.append(u'0') 265 | elif item.is_punct: 266 | out_list.append(u'0') 267 | elif item.is_digit or item.like_num: 268 | out_list.append(u'0') 269 | elif item.like_email: 270 | out_list.append(u'0') 271 | elif item.like_url: 272 | out_list.append(u'0') 273 | elif item.is_stop: 274 | out_list.append(u'0') 275 | else: 276 | out_list.append(item.lemma_) 277 | if location.strip() != u"" and (item.ent_type == 0 or index == len(in_list) - 1): 278 | location = location.strip() 279 | coords = get_coordinates(c, location) 280 | if len(coords) > 0 and location != u" ".join(target): 281 | if is_near: 282 | locations_near.append(coords) 283 | else: 284 | locations_far.append(coords) 285 | else: 286 | offset = 1 if index == len(in_list) - 1 else 0 287 | for i in range(index - len(location.split()), index): 288 | out_list[i + offset] = in_list[i + offset].lemma_ \ 289 | if in_list[i + offset].is_alpha and location != u" ".join(target) else u'0' 290 | location = u"" 291 | target_grid = get_coordinates(c, u" ".join(target)) 292 | if len(target_grid) == 0: 293 | skipped += 1 294 | break 295 | entities_near = merge_lists(locations_near) 296 | entities_far = merge_lists(locations_far) 297 | locations_near, locations_far = [], [] 298 | o.write(lat + u"\t" + lon + u"\t" + str(near_out) + u"\t" + str(far_out) + u"\t") 299 | o.write(str(target_grid) + u"\t" + str([t.lower() for t in target][:TARGET_LENGTH])) 300 | o.write(u"\t" + str(entities_near) + u"\t" + str(entities_far) + u"\n") 301 | limit += 1 302 | if limit > 29: 303 | break 304 | line = line.strip().split("\t") 305 | if u"(" in line[1]: 306 | line[1] = line[1].split(u"(")[0].strip() 307 | if line[1].strip().startswith(u"Geography of "): 308 | target = line[1].replace(u"Geography of ", u"").split() 309 | elif u"," in line[1]: 310 | target = line[1].split(u",")[0].strip().split() 311 | else: 312 | target = line[1].split() 313 | lat = line[2] 314 | lon = line[3] 315 | string = "" 316 | print(u"Processed", limit, u"Skipped:", skipped, u"Name:", u" ".join(target)) 317 | else: 318 | string += line 319 | o.close() 320 | 321 | 322 | def generate_evaluation_data(corpus, file_name): 323 | """ 324 | Create evaluation data from text files. See README for formatting and download instructions. 325 | :param corpus: name of the dataset such as LGL, GEOVIRUS or WIKTOR 326 | :param file_name: an affix, in case you're creating several versions of the same dataset 327 | """ 328 | conn = sqlite3.connect(u'../data/geonames.db') 329 | c = conn.cursor() 330 | nlp = spacy.load(u'en') # or spacy.load(u'en_core_web_lg'), it depends on your choice of model 331 | padding = nlp(u"0")[0] 332 | directory = u"../data/" + corpus + u"/" 333 | o = codecs.open(u"data/eval_" + corpus + file_name + u".txt", u"w", encoding=u"utf-8") 334 | line_no = 0 if corpus == u"lgl" else -1 335 | 336 | for line in codecs.open(u"data/" + corpus + file_name + u".txt", u"r", encoding=u"utf-8"): 337 | line_no += 1 338 | if len(line.strip()) == 0: 339 | continue 340 | for toponym in line.split(u"||")[:-1]: 341 | captured = False 342 | doc = nlp(codecs.open(directory + str(line_no), u"r", encoding=u"utf-8").read()) 343 | locations_near, locations_far = [], [] 344 | toponym = toponym.split(u",,") 345 | target = [t.text for t in nlp(toponym[1])] 346 | ent_length = len(u" ".join(target)) 347 | lat, lon = toponym[2], toponym[3] 348 | start, end = int(toponym[4]), int(toponym[5]) 349 | for d in doc: 350 | if d.text == target[0]: 351 | if u" ".join(target) == u" ".join([t.text for t in doc[d.i:d.i + len(target)]]): 352 | if abs(d.idx - start) > 4 or abs(d.idx + ent_length - end) > 4: 353 | continue 354 | captured = True 355 | near_inp = pad_list(CONTEXT_LENGTH / 2, [x for x in doc[max(0, d.i - CONTEXT_LENGTH / 2):d.i]], True, padding) \ 356 | + pad_list(CONTEXT_LENGTH / 2, [x for x in doc[d.i + len(target): d.i + len(target) + CONTEXT_LENGTH / 2]], False, padding) 357 | far_inp = pad_list(CONTEXT_LENGTH / 2, [x for x in doc[max(0, d.i - CONTEXT_LENGTH):max(0, d.i - CONTEXT_LENGTH / 2)]], True, padding) \ 358 | + pad_list(CONTEXT_LENGTH / 2, [x for x in doc[d.i + len(target) + CONTEXT_LENGTH / 2: d.i + len(target) + CONTEXT_LENGTH]], False, padding) 359 | near_out, far_out = [], [] 360 | location = u"" 361 | for (out_list, in_list, is_near) in [(near_out, near_inp, True), (far_out, far_inp, False)]: 362 | for index, item in enumerate(in_list): 363 | if item.ent_type_ in [u"GPE", u"FACILITY", u"LOC", u"FAC", u"LOCATION"]: 364 | if item.ent_iob_ == u"B" and item.text.lower() == u"the": 365 | out_list.append(item.text.lower()) 366 | else: 367 | location += item.text + u" " 368 | out_list.append(u"**LOC**" + item.text.lower()) 369 | elif item.ent_type_ in [u"PERSON", u"DATE", u"TIME", u"PERCENT", u"MONEY", 370 | u"QUANTITY", u"CARDINAL", u"ORDINAL"]: 371 | out_list.append(u'0') 372 | elif item.is_punct: 373 | out_list.append(u'0') 374 | elif item.is_digit or item.like_num: 375 | out_list.append(u'0') 376 | elif item.like_email: 377 | out_list.append(u'0') 378 | elif item.like_url: 379 | out_list.append(u'0') 380 | elif item.is_stop: 381 | out_list.append(u'0') 382 | else: 383 | out_list.append(item.lemma_) 384 | if location.strip() != u"" and (item.ent_type == 0 or index == len(in_list) - 1): 385 | location = location.strip() 386 | coords = get_coordinates(c, location) 387 | if len(coords) > 0 and location != u" ".join(target): 388 | if is_near: 389 | locations_near.append(coords) 390 | else: 391 | locations_far.append(coords) 392 | else: 393 | offset = 1 if index == len(in_list) - 1 else 0 394 | for i in range(index - len(location.split()), index): 395 | out_list[i + offset] = in_list[i + offset].lemma_ \ 396 | if in_list[i + offset].is_alpha and location != u" ".join(target) else u'0' 397 | location = u"" 398 | 399 | lookup = toponym[0] if corpus != u"wiki" else toponym[1] 400 | target_grid = get_coordinates(c, lookup) 401 | if len(target_grid) == 0: 402 | raise Exception(u"No entry in the database!", lookup) 403 | entities_near = merge_lists(locations_near) 404 | entities_far = merge_lists(locations_far) 405 | locations_near, locations_far = [], [] 406 | o.write(lat + u"\t" + lon + u"\t" + str(near_out) + u"\t" + str(far_out) + u"\t") 407 | o.write(str(target_grid) + u"\t" + str([t.lower() for t in lookup.split()][:TARGET_LENGTH])) 408 | o.write(u"\t" + str(entities_near) + u"\t" + str(entities_far) + u"\n") 409 | if not captured: 410 | print line_no, line, target, start, end 411 | o.close() 412 | 413 | 414 | def visualise_2D_grid(x, title, log=False): 415 | """ 416 | Display 2D array data with a title. Optional: log for better visualisation of small values. 417 | :param x: 2D numpy array you want to visualise 418 | :param title: of the chart because it's nice to have one :-) 419 | :param log: True in order to log the values and make for better visualisation, False for raw numbers 420 | """ 421 | if log: 422 | x = np.log10(x) 423 | cmap = colors.LinearSegmentedColormap.from_list('my_colormap', ['lightgrey', 'darkgrey', 'dimgrey', 'black']) 424 | cmap.set_bad(color='white') 425 | img = pyplot.imshow(x, cmap=cmap, interpolation='nearest') 426 | pyplot.colorbar(img, cmap=cmap) 427 | plt.title(title) 428 | # plt.savefig(title + u".png", dpi=200, transparent=True) # Uncomment to save to file 429 | plt.show() 430 | 431 | 432 | def generate_vocabulary(path, min_words, min_entities): 433 | """ 434 | Prepare the vocabulary for training/testing. This function is to be called on generated data only, not plain text. 435 | :param path: to the file from which to build 436 | :param min_words: occurrence for inclusion in the vocabulary 437 | :param min_entities: occurrence for inclusion in the vocabulary 438 | """ 439 | vocab_words, vocab_locations = {UNKNOWN, u'0'}, {UNKNOWN, u'0'} 440 | words, locations = [], [] 441 | for f in [path]: # You can also build the vocabulary from several files, just add to the list. 442 | training_file = codecs.open(f, u"r", encoding=u"utf-8") 443 | for line in training_file: 444 | line = line.strip().split("\t") 445 | words.extend([w for w in eval(line[2]) if u"**LOC**" not in w]) # NEAR WORDS 446 | words.extend([w for w in eval(line[3]) if u"**LOC**" not in w]) # FAR WORDS 447 | locations.extend([w for w in eval(line[2]) if u"**LOC**" in w]) # NEAR ENTITIES 448 | locations.extend([w for w in eval(line[3]) if u"**LOC**" in w]) # FAR ENTITIES 449 | 450 | words = Counter(words) 451 | for word in words: 452 | if words[word] > min_words: 453 | vocab_words.add(word) 454 | print(u"Words saved:", len(vocab_words)) 455 | 456 | locations = Counter(locations) 457 | for location in locations: 458 | if locations[location] > min_entities: 459 | vocab_locations.add(location.replace(u"**LOC**", u"")) 460 | print(u"Locations saved:", len(vocab_locations)) 461 | 462 | vocabulary = vocab_words.union(vocab_locations) 463 | word_to_index = dict([(w, i) for i, w in enumerate(vocabulary)]) 464 | cPickle.dump(word_to_index, open(u"data/words2index.pkl", "w")) 465 | 466 | 467 | def generate_arrays_from_file(path, words_to_index, train=True): 468 | """ 469 | Generator function for the FULL (SOTA) CNN + map_vector model in the paper. Uses all available data inputs. 470 | :param path: to the training file (see training data generation functions) 471 | :param words_to_index: the vocabulary set 472 | :param train: True is generating training data, false for test data 473 | """ 474 | while True: 475 | training_file = codecs.open(path, "r", encoding="utf-8") 476 | counter = 0 477 | context_words, entities_strings, labels = [], [], [] 478 | map_vector, target_string = [], [] 479 | for line in training_file: 480 | counter += 1 481 | line = line.strip().split("\t") 482 | labels.append(construct_map_vector([(float(line[0]), float(line[1]), 0)], 2, ENCODING_MAP_2x2, OUTLIERS_MAP_2x2)) 483 | 484 | near = [w if u"**LOC**" not in w else u'0' for w in eval(line[2])] 485 | far = [w if u"**LOC**" not in w else u'0' for w in eval(line[3])] 486 | context_words.append(far[:CONTEXT_LENGTH / 2] + near + far[CONTEXT_LENGTH / 2:]) 487 | 488 | near = [w.replace(u"**LOC**", u"") if u"**LOC**" in w else u'0' for w in eval(line[2])] 489 | far = [w.replace(u"**LOC**", u"") if u"**LOC**" in w else u'0' for w in eval(line[3])] 490 | entities_strings.append(far[:CONTEXT_LENGTH / 2] + near + far[CONTEXT_LENGTH / 2:]) 491 | 492 | # map_vector.append(construct_map_vector(sorted(eval(line[4]) + eval(line[6]) + eval(line[7]), 493 | # key=lambda (a, b, c, d): c, reverse=True), 1, ENCODING_MAP_1x1, OUTLIERS_MAP_1x1)) 494 | # paper version above versus small experimental setup below, map_vector is fully modular, remember? Try both! 495 | map_vector.append(construct_map_vector(eval(line[4]) + eval(line[6]) + eval(line[7]), 1, ENCODING_MAP_1x1, OUTLIERS_MAP_1x1)) 496 | target_string.append(pad_list(TARGET_LENGTH, eval(line[5]), True, u'0')) 497 | 498 | if counter % BATCH_SIZE == 0: 499 | for collection in [context_words, entities_strings, target_string]: 500 | for x in collection: 501 | for i, w in enumerate(x): 502 | if w in words_to_index: 503 | x[i] = words_to_index[w] 504 | else: 505 | x[i] = words_to_index[UNKNOWN] 506 | if train: 507 | yield ([np.asarray(context_words), np.asarray(context_words), np.asarray(entities_strings), 508 | np.asarray(entities_strings), np.asarray(map_vector), np.asarray(target_string)], np.asarray(labels)) 509 | else: 510 | yield ([np.asarray(context_words), np.asarray(context_words), np.asarray(entities_strings), 511 | np.asarray(entities_strings), np.asarray(map_vector), np.asarray(target_string)]) 512 | 513 | context_words, entities_strings, labels = [], [], [] 514 | map_vector, target_string = [], [] 515 | 516 | if len(labels) > 0: # This block is only ever entered at the end to yield the final few samples. (< BATCH_SIZE) 517 | for collection in [context_words, entities_strings, target_string]: 518 | for x in collection: 519 | for i, w in enumerate(x): 520 | if w in words_to_index: 521 | x[i] = words_to_index[w] 522 | else: 523 | x[i] = words_to_index[UNKNOWN] 524 | if train: 525 | yield ([np.asarray(context_words), np.asarray(context_words), np.asarray(entities_strings), 526 | np.asarray(entities_strings), np.asarray(map_vector), np.asarray(target_string)], np.asarray(labels)) 527 | else: 528 | yield ([np.asarray(context_words), np.asarray(context_words), np.asarray(entities_strings), 529 | np.asarray(entities_strings), np.asarray(map_vector), np.asarray(target_string)]) 530 | 531 | 532 | def generate_arrays_from_file_lstm(path, words_to_index, train=True): 533 | """ 534 | Generator for the context2vec model. Uses only lexical features. 535 | To replicate the map_vector + CONTEXT2VEC model from the paper, uncomment a few sections below 536 | and in the context2vec.py file. I hope it's clear enough :-) Email me if it isn't! 537 | :param path: to the training file (see training data generation functions) 538 | :param words_to_index: the vocabulary set 539 | :param train: True for training stage, False for testing stage 540 | """ 541 | while True: 542 | training_file = codecs.open(path, "r", encoding="utf-8") 543 | counter = 0 544 | left, right, map_vector = [], [], [] 545 | target_string, labels = [], [] 546 | for line in training_file: 547 | counter += 1 548 | line = line.strip().split("\t") 549 | labels.append(construct_map_vector([(float(line[0]), float(line[1]), 0)], 2, ENCODING_MAP_2x2, OUTLIERS_MAP_2x2)) 550 | 551 | near = [w.replace(u"**LOC**", u"") for w in eval(line[2])] 552 | far = [w.replace(u"**LOC**", u"") for w in eval(line[3])] 553 | left.append(far[:CONTEXT_LENGTH / 2] + near[:CONTEXT_LENGTH / 2]) 554 | right.append(near[CONTEXT_LENGTH / 2:] + far[CONTEXT_LENGTH / 2:]) 555 | 556 | target_string.append(pad_list(TARGET_LENGTH, eval(line[5]), True, u'0')) 557 | 558 | # map_vector.append(construct_map_vector(eval(line[4]) + eval(line[6]) + eval(line[7]), 1, ENCODING_MAP_1x1, OUTLIERS_MAP_1x1)) 559 | 560 | if counter % BATCH_SIZE == 0: 561 | for collection in [left, right, target_string]: 562 | for x in collection: 563 | for i, w in enumerate(x): 564 | if w in words_to_index: 565 | x[i] = words_to_index[w] 566 | else: 567 | x[i] = words_to_index[UNKNOWN] 568 | if train: 569 | yield ([np.asarray(left), np.asarray(right), np.asarray(target_string)], np.asarray(labels)) 570 | # yield ([np.asarray(left), np.asarray(right), np.asarray(map_vector), np.asarray(target_string)], np.asarray(labels)) 571 | else: 572 | yield ([np.asarray(left), np.asarray(right), np.asarray(target_string)]) 573 | # yield ([np.asarray(left), np.asarray(right), np.asarray(map_vector), np.asarray(target_string)]) 574 | 575 | left, right, map_vector = [], [], [] 576 | target_string, labels = [], [] 577 | 578 | if len(labels) > 0: # This block is only ever entered at the end to yield the final few samples. (< BATCH_SIZE) 579 | for collection in [left, right, target_string]: 580 | for x in collection: 581 | for i, w in enumerate(x): 582 | if w in words_to_index: 583 | x[i] = words_to_index[w] 584 | else: 585 | x[i] = words_to_index[UNKNOWN] 586 | if train: 587 | yield ([np.asarray(left), np.asarray(right), np.asarray(target_string)], np.asarray(labels)) 588 | # yield ([np.asarray(left), np.asarray(right), np.asarray(map_vector), np.asarray(target_string)], np.asarray(labels)) 589 | else: 590 | yield ([np.asarray(left), np.asarray(right), np.asarray(target_string)]) 591 | # yield ([np.asarray(left), np.asarray(right), np.asarray(map_vector), np.asarray(target_string)]) 592 | 593 | 594 | def generate_strings_from_file(path): 595 | """ 596 | Generator of labels, location names and context. Used for training and testing. 597 | :param path: to the training file (see training data generation functions) 598 | :return: Yields a list of tuples [(label, location name, context), ...] 599 | """ 600 | while True: 601 | for line in codecs.open(path, "r", encoding="utf-8"): 602 | line = line.strip().split("\t") 603 | context = u" ".join(eval(line[2])) + u"*E*" + u" ".join(eval(line[5])) + u"*E*" + u" ".join(eval(line[3])) 604 | yield ((float(line[0]), float(line[1])), u" ".join(eval(line[5])).strip(), context) 605 | 606 | 607 | def generate_arrays_from_file_map_vector(path, train=True, looping=True): 608 | """ 609 | Generator for the plain map_vector model, works for MLP, Naive Bayes or Random Forest. Table 2 in the paper. 610 | :param path: to the training file (see training data generation functions) 611 | :param train: True for training phase, False for testing phase 612 | :param looping: True for continuous generation, False for one iteration. 613 | """ 614 | while True: 615 | training_file = codecs.open(path, "r", encoding="utf-8") 616 | counter = 0 617 | labels, target_coord = [], [] 618 | for line in training_file: 619 | counter += 1 620 | line = line.strip().split("\t") 621 | labels.append(construct_map_vector([(float(line[0]), float(line[1]), 0, u'')], 2, ENCODING_MAP_2x2, OUTLIERS_MAP_2x2)) 622 | target_coord.append(construct_map_vector(eval(line[4]) + eval(line[6]) + eval(line[7]), 1, ENCODING_MAP_1x1, OUTLIERS_MAP_1x1)) 623 | 624 | if counter % BATCH_SIZE == 0: 625 | if train: 626 | yield ([np.asarray(target_coord)], np.asarray(labels)) 627 | else: 628 | yield ([np.asarray(target_coord)]) 629 | 630 | labels = [] 631 | target_coord = [] 632 | 633 | if len(labels) > 0: 634 | # This block is only ever entered at the end to yield the final few samples. (< BATCH_SIZE) 635 | if train: 636 | yield ([np.asarray(target_coord)], np.asarray(labels)) 637 | else: 638 | yield ([np.asarray(target_coord)]) 639 | if not looping: 640 | break 641 | 642 | 643 | def shrink_map_vector(polygon_size): 644 | """ 645 | Remove polygons that only cover oceans. Dumps a dictionary of DB entries. 646 | :param polygon_size: the size of each polygon such as 1x1 or 2x2 or 3x3 degrees (integer) 647 | """ 648 | map_vector = np.zeros((180 / polygon_size) * (360 / polygon_size),) 649 | for line in codecs.open(u"../data/allCountries.txt", u"r", encoding=u"utf-8"): 650 | line = line.split("\t") 651 | lat, lon = float(line[4]), float(line[5]) 652 | index = coord_to_index((lat, lon), polygon_size=polygon_size) 653 | map_vector[index] += 1.0 654 | cPickle.dump(map_vector, open(u"mapvec_shrink.pkl", "w")) 655 | 656 | 657 | def oracle(path): 658 | """ 659 | Calculate the Oracle (best possible given your database) performance for a given dataset. 660 | Prints the Oracle scores including mean, median, AUC and acc@161. 661 | :param path: file path to evaluate 662 | """ 663 | final_errors = [] 664 | conn = sqlite3.connect(u'../data/geonames.db') 665 | for line in codecs.open(path, "r", encoding="utf-8"): 666 | line = line.strip().split("\t") 667 | coordinates = (float(line[0]), float(line[1])) 668 | best_candidate = [] 669 | for candidate in get_coordinates(conn.cursor(), u" ".join(eval(line[5])).strip()): 670 | best_candidate.append(great_circle(coordinates, (float(candidate[0]), float(candidate[1]))).km) 671 | final_errors.append(sorted(best_candidate)[0]) 672 | print_stats(final_errors) 673 | 674 | 675 | # --------------------------------------------- INVOKE FUNCTIONS --------------------------------------------------- 676 | # prepare_geocorpora() 677 | # print get_coordinates(sqlite3.connect('../data/geonames.db').cursor(), u"dublin") 678 | # generate_training_data() 679 | # generate_evaluation_data(corpus="geovirus", file_name="") 680 | # generate_vocabulary(path=u"../data/train_wiki.txt", min_words=9, min_entities=1) 681 | # shrink_map_vector(2) 682 | # oracle(u"data/eval_geovirus_gold.txt") 683 | # conn = sqlite3.connect('../data/geonames.db') 684 | # c = conn.cursor() 685 | # c.execute("INSERT INTO GEO VALUES (?, ?)", (u"darfur", u"[(13.5, 23.5, 0), (44.05135, -94.83804, 106)]")) 686 | # c.execute("DELETE FROM GEO WHERE name = 'darfur'") 687 | # conn.commit() 688 | # print index_to_coord(8177, 2) 689 | # populate_sql() 690 | 691 | # -------- CREATE MAPS (mapping from 64,000/16,200 polygons to 23,002, 7,821) ------------ 692 | # map_vector = list(cPickle.load(open(u"data/1x1_geonames.pkl"))) 693 | # zeros = dict([(i, v) for i, v in enumerate(map_vector) if v > 0]) # isolate the non zero values 694 | # zeros = dict([(i, v) for i, v in enumerate(zeros)]) # replace counts with indices 695 | # zeros = dict([(v, i) for (i, v) in zeros.iteritems()]) # reverse keys and values 696 | # cPickle.dump(zeros, open(u"data/1x1_encode_map.pkl", "w")) 697 | 698 | # ------- VISUALISE THE WHOLE DATABASE ---------- 699 | # map_vector = np.reshape(map_vector, newshape=((180 / 1), (360 / 1))) 700 | # visualise_2D_grid(map_vector, "Geonames Database", True) 701 | 702 | # -------- CREATE OUTLIERS (polygons outside of map_vector) MAP -------- 703 | # filtered = [i for i, v in enumerate(map_vector) if v > 0] 704 | # the_rest = [i for i, v in enumerate(map_vector) if v == 0] 705 | # poly_size = 2 706 | # dict_rest = dict() 707 | # 708 | # for poly_rest in the_rest: 709 | # best_index = 100000 710 | # best_dist = 100000 711 | # for poly_filtered in filtered: 712 | # dist = great_circle(index_to_coord(poly_rest, poly_size), index_to_coord(poly_filtered, poly_size)).km 713 | # if dist < best_dist: 714 | # best_index = poly_filtered 715 | # best_dist = dist 716 | # dict_rest[poly_rest] = best_index 717 | # 718 | # cPickle.dump(dict_rest, open(u"data/2x2_outliers_map.pkl", "w")) 719 | 720 | # ------ PROFILING SETUP ----------- 721 | # import cProfile, pstats, StringIO 722 | # pr = cProfile.Profile() 723 | # pr.enable() 724 | # CODE HERE 725 | # pr.disable() 726 | # s = StringIO.StringIO() 727 | # sortby = 'cumulative' 728 | # ps = pstats.Stats(pr, stream=s).sort_stats(sortby) 729 | # ps.print_stats() 730 | # print s.getvalue() 731 | 732 | # ----------- VISUALISATION OF DIFFERENT LOCATIONS ------------- 733 | # print len(get_coordinates(sqlite3.connect('../data/geonames.db').cursor(), u"Melbourne")) 734 | # coord = get_coordinates(sqlite3.connect('../data/geonames.db').cursor(), u"Giza") 735 | # print coord 736 | # coord.extend(get_coordinates(sqlite3.connect('../data/geonames.db').cursor(), u"Giza Plateau")) 737 | # coord.extend(get_coordinates(sqlite3.connect('../data/geonames.db').cursor(), u"Cairo")) 738 | # coord.extend(get_coordinates(sqlite3.connect('../data/geonames.db').cursor(), u"Egypt")) 739 | # coord = sorted(coord, key=lambda (a, b, c, d): c, reverse=True) 740 | # x = construct_map_vector_full_scale(coord, polygon_size=2) 741 | # x = np.reshape(x, newshape=((180 / 2), (360 / 2))) 742 | # visualise_2D_grid(x, "Giza, Giza Plateau, Egypt, Cairo", True) 743 | 744 | # ---------- DUMP DATABASE ------ 745 | # import sqlite3 746 | # 747 | # con = sqlite3.connect('../data/geonames.db') 748 | # with codecs.open('dump.sql', 'w', 'utf-8') as f: 749 | # for line in con.iterdump(): 750 | # f.write('%s\n' % line) 751 | # ------------------------------- 752 | -------------------------------------------------------------------------------- /subsample.py: -------------------------------------------------------------------------------- 1 | import codecs 2 | import sqlite3 3 | from geopy.distance import great_circle 4 | from preprocessing import get_coordinates 5 | 6 | counter = 0 # keeps track of current line number 7 | start = 0 # where do you want to start sampling from? 8 | finish = 800000 # where do you want to end the uniform sampling? 9 | frequency = 2 # 1 means take EVERY sample, 2 means take every SECOND sample, etc... 10 | output_file = u"../data/train_wiki_uniform.txt" # This file is used in train.py 11 | input_file = u"../data/train_wiki.txt" # This dataset contains around 1.4M lines of train examples 12 | 13 | filtering = True # Do you want to filter samples with coordinate errors? Probably yes. 14 | filtered_count = 0 # Keeping track of how many get filtered out? Good idea. 15 | saved_count = 0 # Keeping track of how many samples were saved? That, too. 16 | max_distance = 999 # The maximum size of the coordinate error, this depends on the database. 999 is good. 17 | conn = sqlite3.connect(u'../data/geonames.db') # Download this file from GitHub (milangritta) 18 | c = conn.cursor() # Initialise database connection 19 | 20 | out = codecs.open(output_file, u"w", encoding=u"utf-8") 21 | for line in codecs.open(input_file, u"r", encoding=u"utf-8"): 22 | counter += 1 23 | if counter < start: 24 | continue 25 | if counter > finish: 26 | break 27 | if counter % frequency == 0: 28 | if not filtering: 29 | out.write(line) 30 | saved_count += 1 31 | else: 32 | split = line.split(u"\t") 33 | wiki_coordinates = (float(split[0]), float(split[1])) 34 | name = u" ".join(eval(split[5])).strip() 35 | db_coordinates = get_coordinates(c, name) 36 | distance = [] 37 | for candidate in db_coordinates: 38 | distance.append(great_circle(wiki_coordinates, (float(candidate[0]), float(candidate[1]))).kilometers) 39 | distance = sorted(distance) 40 | if distance[0] > max_distance: 41 | print(name, distance[0]) 42 | filtered_count += 1 43 | else: 44 | out.write(line) 45 | saved_count += 1 46 | 47 | print(u"Saved", saved_count, u"samples.") 48 | if filtering: 49 | print(u"Filtered:", filtered_count, u"samples.") 50 | -------------------------------------------------------------------------------- /test.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | import numpy as np 3 | import cPickle 4 | import sqlite3 5 | import sys 6 | from geopy.distance import great_circle 7 | from keras.models import load_model 8 | from subprocess import check_output 9 | from preprocessing import get_coordinates, print_stats, index_to_coord, generate_strings_from_file 10 | from preprocessing import BATCH_SIZE, REVERSE_MAP_2x2 11 | from preprocessing import generate_arrays_from_file 12 | 13 | # For command line use, type: python test.py 14 | # For example: python test.py lgl_gold 15 | if len(sys.argv) > 1: 16 | test_data = sys.argv[1] 17 | else: 18 | test_data = u"geovirus" # or edit this line if running inside an IDE editor 19 | 20 | saved_model_file = u"../data/weights" 21 | print(u"Testing:", test_data, u"with weights:", saved_model_file) 22 | word_to_index = cPickle.load(open(u"data/words2index.pkl")) # This is the vocabulary file 23 | # -------------------------------------------------------------------------------------------------------------------- 24 | print(u'Loading model...') 25 | model = load_model(saved_model_file) 26 | print(u'Finished loading model...') 27 | # -------------------------------------------------------------------------------------------------------------------- 28 | print(u'Crunching numbers, sit tight...') 29 | # errors = codecs.open(u"errors.tsv", u"w", encoding=u"utf-8") 30 | # Uncomment the above line for error diagnostics, also the section below. 31 | conn = sqlite3.connect(u'../data/geonames.db') 32 | file_name = u"data/eval_" + test_data + u".txt" 33 | final_errors = [] 34 | for prediction, (y, name, context) in zip(model.predict_generator(generate_arrays_from_file(file_name, word_to_index, train=False), 35 | steps=int(check_output([u"wc", file_name]).split()[0]) / BATCH_SIZE, verbose=True), generate_strings_from_file(file_name)): 36 | prediction = index_to_coord(REVERSE_MAP_2x2[np.argmax(prediction)], 2) 37 | candidates = get_coordinates(conn.cursor(), name) 38 | 39 | if len(candidates) == 0: 40 | print(u"Don't have an entry for", name, u"in GeoNames") 41 | raise Exception(u"Check your database, buddy :-)") 42 | 43 | # candidates = [candidates[0]] # Uncomment for population heuristic. 44 | # THE ABOVE IS THE POPULATION ONLY BASELINE IMPLEMENTATION 45 | 46 | best_candidate = [] 47 | max_pop = candidates[0][2] 48 | bias = 0.905 # the Bias parameter in the paper 49 | for candidate in candidates: 50 | err = great_circle(prediction, (float(candidate[0]), float(candidate[1]))).km 51 | best_candidate.append((err - (err * max(1, candidate[2]) / max(1, max_pop)) * bias, (float(candidate[0]), float(candidate[1])))) 52 | best_candidate = sorted(best_candidate, key=lambda (a, b): a)[0] 53 | final_errors.append(great_circle(best_candidate[1], y).km) 54 | 55 | # ---------------- ERROR DIAGNOSTICS -------------------- 56 | # dist_p, dist_y, index_p, index_y = 100000, 100000, 0, 0 57 | # for index, candidate in enumerate(candidates): 58 | # if great_circle(best_candidate[1], (candidate[0], candidate[1])).km < dist_p: 59 | # dist_p = great_circle(best_candidate[1], (candidate[0], candidate[1])).km 60 | # index_p = index 61 | # if great_circle(y, (candidate[0], candidate[1])).km < dist_y: 62 | # dist_y = great_circle(y, (candidate[0], candidate[1])).km 63 | # index_y = index 64 | # 65 | # errors.write(name + u"\t" + unicode(y) + u"\t" + unicode(p) + "\t" + unicode(best_candidate[1]) 66 | # + u"\t" + unicode(index_y) + u"\t" + unicode(index_p) + u"\t" + unicode(final_errors[-1]) + u"\t" + 67 | # unicode(best_candidate[0]) + u"\t" + unicode(len(candidates)) + u"\t" + context.replace(u"\n", u"") + u"\n") 68 | # ------------------ END OF DIAGNOSTICS ----------------- 69 | 70 | print_stats(final_errors) 71 | print(u"Processed file", file_name) 72 | 73 | # ------------------------ VISUALISATION ------------------------------ 74 | # import matplotlib.pyplot as plt 75 | # plt.plot(range(len(final_errors)), np.log(1 + np.asarray(sorted(final_errors)))) 76 | # plt.xlabel(u"Predictions") 77 | # plt.ylabel(u'Error Size') 78 | # plt.title(u"Some Chart") 79 | # plt.savefig(u'test.png', transparent=True) 80 | # plt.show() 81 | # ---------------------------------------------------------------------- 82 | -------------------------------------------------------------------------------- /text2mapVec.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | import sqlite3 3 | import cPickle 4 | import numpy as np 5 | import spacy 6 | 7 | ####################################################################################### 8 | # # 9 | # If you're only interested the Map Vector generation, I extracted the relevant # 10 | # code into this python script for quick and dirty use. You still need at least the # 11 | # encoding map files (see ENCODING_MAP below). You also need a database 'geonames.db' # 12 | # # 13 | ####################################################################################### 14 | 15 | 16 | def text2mapvec(doc, mapping, outliers, polygon_size, db, exclude): 17 | """ 18 | Parse text, extract entities, create and return the MAP VECTOR. 19 | :param exclude: Exclude the TARGET entity from the MAP VECTOR to avoid self-bias 20 | :param db: database cursor to use for retrieval 21 | :param doc: the paragraph to turn into a Map Vector 22 | :param mapping: the map resolution file, determines the size of MAP VECTOR 23 | :param outliers: must be the same size/resolution as MAPPING 24 | :param polygon_size: the tile size must also match i.e. all three either 1x1 or 2x2, etc. 25 | :return: the map vector for this paragraph of text 26 | """ 27 | location = u"" 28 | entities = [] 29 | for index, item in enumerate(doc): 30 | if item.ent_type_ in [u"GPE", u"FACILITY", u"LOC", u"FAC", u"LOCATION"]: 31 | if not (item.ent_iob_ == u"B" and item.text.lower() == u"the"): 32 | location += item.text + u" " 33 | 34 | if location.strip() != u"" and (item.ent_type == 0 or index == len(doc) - 1): 35 | location = location.strip() 36 | if exclude is not None: 37 | if location == exclude: 38 | location = u"" 39 | continue 40 | coords = get_coordinates(db, location) 41 | if len(coords) > 0: 42 | entities.extend(coords) 43 | location = u"" 44 | 45 | entities = sorted(entities, key=lambda (a, b, c, d): c, reverse=True) 46 | mapvec = np.zeros(len(mapping), ) 47 | 48 | if len(entities) == 0: 49 | return mapvec # No locations? Return an empty vector. 50 | max_pop = entities[0][2] if entities[0][2] > 0 else 1 51 | for s in entities: 52 | index = coord_to_index((s[0], s[1]), polygon_size) 53 | if index in mapping: 54 | index = mapping[index] 55 | else: 56 | index = mapping[outliers[index]] 57 | mapvec[index] += float(max(s[2], 1)) / max_pop 58 | return mapvec / mapvec.max() if mapvec.max() > 0.0 else mapvec 59 | 60 | 61 | def coord_to_index(coordinates, polygon_size): 62 | """ 63 | Convert coordinates into an array index. Use that to modify mapvec polygon value. 64 | :param coordinates: (latitude, longitude) to compute 65 | :param polygon_size: integer size of the polygon? i.e. the resolution of the world 66 | :return: index pointing into mapvec array 67 | """ 68 | latitude = float(coordinates[0]) - 90 if float(coordinates[0]) != -90 else -179.99 # The two edge cases must 69 | longitude = float(coordinates[1]) + 180 if float(coordinates[1]) != 180 else 359.99 # get handled differently! 70 | if longitude < 0: 71 | longitude = -longitude 72 | if latitude < 0: 73 | latitude = -latitude 74 | x = int(360 / polygon_size) * int(latitude / polygon_size) 75 | y = int(longitude / polygon_size) 76 | return x + y if 0 <= x + y <= int(360 / polygon_size) * int(180 / polygon_size) else Exception(u"Shock horror!!") 77 | 78 | 79 | def get_coordinates(con, loc_name): 80 | """ 81 | Access the database to retrieve coordinates and other data from DB. 82 | :param con: sqlite3 database cursor i.e. DB connection 83 | :param loc_name: name of the place 84 | :return: a list of tuples [(latitude, longitude, population, feature_code), ...] 85 | """ 86 | result = con.execute(u"SELECT METADATA FROM GEO WHERE NAME = ?", (loc_name.lower(),)).fetchone() 87 | if result: 88 | result = eval(result[0]) # Do not remove the sorting, the function below assumes sorted results! 89 | return sorted(result, key=lambda (a, b, c, d): c, reverse=True) 90 | else: 91 | return [] 92 | 93 | 94 | def buildMapVec(text): 95 | """ 96 | An example wrapper function for text2mapVec(), reads in necessary collections and then runs text2mapVec(). 97 | Feel free to modify to your preference and task objective. 98 | :param text: to create the Map Vector from encoded as unicode. 99 | :return: currently only prints the vector, add 'return map_vector' or whatever you prefer. 100 | """ 101 | ENCODING_MAP = cPickle.load(open(u"data/1x1_encode_map.pkl")) # the resolution of the map 102 | OUTLIERS_MAP = cPickle.load(open(u"data/1x1_outliers_map.pkl")) # dimensions must match the above 103 | nlp = spacy.load(u'en_core_web_lg') # or spacy.load(u'en') depending on your Spacy Download (simple or full) 104 | conn = sqlite3.connect(u'../data/geonames.db').cursor() # this DB can be downloaded using the GitHub link 105 | map_vector = text2mapvec(doc=nlp(text), mapping=ENCODING_MAP, outliers=OUTLIERS_MAP, polygon_size=1, db=conn, exclude=u"Cairo") 106 | print(map_vector) 107 | 108 | 109 | # text = u"The Giza pyramid complex is an archaeological site on the Giza Plateau, on the outskirts of Cairo, Egypt." 110 | # buildMapVec(text) 111 | -------------------------------------------------------------------------------- /train.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | import codecs 3 | import numpy as np 4 | import cPickle 5 | from keras import Input 6 | from keras.callbacks import ModelCheckpoint, EarlyStopping 7 | from keras.engine import Model 8 | from keras.layers.merge import concatenate 9 | from keras.layers import Embedding, Dense, Dropout, Conv1D, GlobalMaxPooling1D 10 | from preprocessing import BATCH_SIZE, EMBEDDING_DIMENSION, CONTEXT_LENGTH, UNKNOWN 11 | from preprocessing import TARGET_LENGTH, generate_arrays_from_file, ENCODING_MAP_2x2, ENCODING_MAP_1x1 12 | from subprocess import check_output 13 | 14 | print(u"Embedding Dimension:", EMBEDDING_DIMENSION) 15 | print(u"Input length (each side):", CONTEXT_LENGTH) 16 | word_to_index = cPickle.load(open(u"data/words2index.pkl")) 17 | print(u"Vocabulary Size:", len(word_to_index)) 18 | 19 | vectors = {UNKNOWN: np.ones(EMBEDDING_DIMENSION), u'0': np.ones(EMBEDDING_DIMENSION)} 20 | for line in codecs.open(u"../data/glove.twitter." + str(EMBEDDING_DIMENSION) + u"d.txt", encoding=u"utf-8"): 21 | if line.strip() == "": 22 | continue 23 | t = line.split() 24 | vectors[t[0]] = [float(x) for x in t[1:]] 25 | print(u'Vectors...', len(vectors)) 26 | 27 | emb_weights = np.zeros((len(word_to_index), EMBEDDING_DIMENSION)) 28 | oov = 0 29 | for w in word_to_index: 30 | if w in vectors: 31 | emb_weights[word_to_index[w]] = vectors[w] 32 | else: 33 | emb_weights[word_to_index[w]] = np.random.normal(size=(EMBEDDING_DIMENSION,), scale=0.3) 34 | oov += 1 35 | 36 | emb_weights = np.array([emb_weights]) 37 | print(u'Done preparing vectors...') 38 | print(u"OOV (no vectors):", oov) 39 | # -------------------------------------------------------------------------------------------------------------------- 40 | print(u'Building model...') 41 | embeddings = Embedding(len(word_to_index), EMBEDDING_DIMENSION, input_length=CONTEXT_LENGTH * 2, weights=emb_weights) 42 | # shared embeddings between all language input layers 43 | 44 | context_words_pair = Input(shape=(CONTEXT_LENGTH * 2,)) 45 | cwp = embeddings(context_words_pair) 46 | cwp = Conv1D(1000, 2, activation='relu', strides=1)(cwp) 47 | cwp = GlobalMaxPooling1D()(cwp) 48 | cwp = Dense(250)(cwp) 49 | cwp = Dropout(0.5)(cwp) 50 | 51 | context_words_single = Input(shape=(CONTEXT_LENGTH * 2,)) 52 | cws = embeddings(context_words_single) 53 | cws = Conv1D(1000, 1, activation='relu', strides=1)(cws) 54 | cws = GlobalMaxPooling1D()(cws) 55 | cws = Dense(250)(cws) 56 | cws = Dropout(0.5)(cws) 57 | 58 | entities_strings_pair = Input(shape=(CONTEXT_LENGTH * 2,)) 59 | esp = embeddings(entities_strings_pair) 60 | esp = Conv1D(1000, 2, activation='relu', strides=1)(esp) 61 | esp = GlobalMaxPooling1D()(esp) 62 | esp = Dense(250)(esp) 63 | esp = Dropout(0.5)(esp) 64 | 65 | entities_strings_single = Input(shape=(CONTEXT_LENGTH * 2,)) 66 | ess = embeddings(entities_strings_single) 67 | ess = Conv1D(1000, 1, activation='relu', strides=1)(ess) 68 | ess = GlobalMaxPooling1D()(ess) 69 | ess = Dense(250)(ess) 70 | ess = Dropout(0.5)(ess) 71 | 72 | mapvec = Input(shape=(len(ENCODING_MAP_1x1),)) 73 | l2v = Dense(5000, activation='relu', input_dim=len(ENCODING_MAP_1x1))(mapvec) 74 | l2v = Dense(1000, activation='relu')(l2v) 75 | l2v = Dropout(0.5)(l2v) 76 | 77 | target_string = Input(shape=(TARGET_LENGTH,)) 78 | ts = Embedding(len(word_to_index), EMBEDDING_DIMENSION, input_length=TARGET_LENGTH, weights=emb_weights)(target_string) 79 | ts = Conv1D(1000, 3, activation='relu')(ts) 80 | ts = GlobalMaxPooling1D()(ts) 81 | ts = Dropout(0.5)(ts) 82 | 83 | inp = concatenate([cwp, cws, esp, ess, l2v, ts]) 84 | inp = Dense(units=len(ENCODING_MAP_2x2), activation=u'softmax')(inp) 85 | model = Model(inputs=[context_words_pair, context_words_single, entities_strings_pair, entities_strings_single, 86 | mapvec, target_string], outputs=[inp]) 87 | model.compile(loss=u'categorical_crossentropy', optimizer=u'rmsprop', metrics=[u'accuracy']) 88 | 89 | print(u'Finished building model...') 90 | # -------------------------------------------------------------------------------------------------------------------- 91 | checkpoint = ModelCheckpoint(filepath=u"../data/weights.{epoch:02d}-{acc:.2f}.hdf5", verbose=0) 92 | early_stop = EarlyStopping(monitor=u'acc', patience=5) 93 | file_name = u"../data/train_wiki_uniform.txt" 94 | model.fit_generator(generate_arrays_from_file(file_name, word_to_index), 95 | steps_per_epoch=int(check_output(["wc", file_name]).split()[0]) / BATCH_SIZE, 96 | epochs=250, callbacks=[checkpoint, early_stop]) 97 | --------------------------------------------------------------------------------