Datasets:

Modalities:
Text
Formats:
json
ArXiv:
Libraries:
Datasets
pandas
License:
danieldeutsch commited on
Commit
48c4dd6
·
verified ·
1 Parent(s): 11e0d7d

Updates the README

Browse files
Files changed (1) hide show
  1. README.md +154 -0
README.md CHANGED
@@ -275,3 +275,157 @@ configs:
275
  - split: train
276
  path: "en-zu_ZA.jsonl"
277
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
275
  - split: train
276
  path: "en-zu_ZA.jsonl"
277
  ---
278
+
279
+ # WMT24++
280
+
281
+ This repository contains the human translation and post-edit data for the 55 en->xx language pairs released in
282
+ the publication
283
+ [WMT24++: Expanding the Language Coverage of WMT24 to 55 Languages & Dialects]().
284
+ If you are interested in the MT/LLM system outputs and automatic metric scores, please see [MTME](https://github.com/google-research/mt-metrics-eval).
285
+ If you are interested in the images of the source URLs for each document, please see [here](https://huggingface.co/datasets/google/wmt24pp-images).
286
+
287
+ ## Schema
288
+
289
+ Each language pair is stored in its own jsonl file.
290
+ Each row is a serialized JSON object with the following fields:
291
+
292
+ - `lp`: The language pair (e.g., `"en-de_DE"`).
293
+ - `domain`: The domain of the source, either `"canary"`, `"news"`, `"social"`, `"speech"`, or `"literary"`.
294
+ - `document_id`: The unique ID that identifies the document the source came from.
295
+ - `segment_id`: The globally unique ID that identifies the segment.
296
+ - `is_bad_source`: A Boolean that indicates whether this source is low quality (e.g., HTML, URLs, emoijs). In the paper, the segments marked as true were removed from the evaluation, and we recommend doing the same.
297
+ - `source`: The English source text.
298
+ - `target`: The post-edit of `original_target`. We recommend using the post-edit as the default reference.
299
+ - `original_target`: The original reference translation.
300
+
301
+ ## Helpful Python Constants
302
+
303
+ ```python
304
+ LANGUAGE_PAIRS = (
305
+ "en-ar_EG", "en-ar_SA", "en-bg_BG", "en-bn_IN", "en-ca_ES", "en-cs_CZ", "en-da_DK", "en-de_DE",
306
+ "en-el_GR", "en-es_MX", "en-et_EE", "en-fa_IR", "en-fi_FI", "en-fil_PH", "en-fr_CA", "en-fr_FR",
307
+ "en-gu_IN", "en-he_IL", "en-hi_IN", "en-hr_HR", "en-hu_HU", "en-id_ID", "en-is_IS", "en-it_IT",
308
+ "en-ja_JP", "en-kn_IN", "en-ko_KR", "en-lt_LT", "en-lv_LV", "en-ml_IN", "en-mr_IN", "en-nl_NL",
309
+ "en-no_NO", "en-pa_IN", "en-pl_PL", "en-pt_BR", "en-pt_PT", "en-ro_RO", "en-ru_RU", "en-sk_SK",
310
+ "en-sl_SI", "en-sr_RS", "en-sv_SE", "en-sw_KE", "en-sw_TZ", "en-ta_IN", "en-te_IN", "en-th_TH",
311
+ "en-tr_TR", "en-uk_UA", "en-ur_PK", "en-vi_VN", "en-zh_CN", "en-zh_TW", "en-zu_ZA",
312
+ )
313
+
314
+ LANGUAGE_BY_CODE = {
315
+ "ar_EG": "Arabic",
316
+ "ar_SA": "Arabic",
317
+ "bg_BG": "Bulgarian",
318
+ "bn_BD": "Bengali",
319
+ "bn_IN": "Bengali",
320
+ "ca_ES": "Catalan",
321
+ "cs_CZ": "Czech",
322
+ "da_DK": "Danish",
323
+ "de_DE": "German",
324
+ "el_GR": "Greek",
325
+ "es_MX": "Spanish",
326
+ "et_EE": "Estonian",
327
+ "fa_IR": "Farsi",
328
+ "fi_FI": "Finnish",
329
+ "fil_PH": "Filipino",
330
+ "fr_CA": "French",
331
+ "fr_FR": "French",
332
+ "gu_IN": "Gujarati",
333
+ "he_IL": "Hebrew",
334
+ "hi_IN": "Hindi",
335
+ "hr_HR": "Croatian",
336
+ "hu_HU": "Hungarian",
337
+ "id_ID": "Indonesian",
338
+ "is_IS": "Icelandic",
339
+ "it_IT": "Italian",
340
+ "ja_JP": "Japanese",
341
+ "kn_IN": "Kannada",
342
+ "ko_KR": "Korean",
343
+ "lt_LT": "Lithuanian",
344
+ "lv_LV": "Latvian",
345
+ "ml_IN": "Malayalam",
346
+ "mr_IN": "Marathi",
347
+ "nl_NL": "Dutch",
348
+ "no_NO": "Norwegian",
349
+ "pa_IN": "Punjabi",
350
+ "pl_PL": "Polish",
351
+ "pt_BR": "Portuguese",
352
+ "pt_PT": "Portuguese",
353
+ "ro_RO": "Romanian",
354
+ "ru_RU": "Russian",
355
+ "sk_SK": "Slovak",
356
+ "sl_SI": "Slovenian",
357
+ "sr_RS": "Serbian",
358
+ "sv_SE": "Swedish",
359
+ "sw_KE": "Swahili",
360
+ "sw_TZ": "Swahili",
361
+ "ta_IN": "Tamil",
362
+ "te_IN": "Telugu",
363
+ "th_TH": "Thai",
364
+ "tr_TR": "Turkish",
365
+ "uk_UA": "Ukrainian",
366
+ "ur_PK": "Urdu",
367
+ "vi_VN": "Vietnamese",
368
+ "zh_CN": "Mandarin",
369
+ "zh_TW": "Mandarin",
370
+ "zu_ZA": "Zulu",
371
+ }
372
+
373
+ REGION_BY_CODE = {
374
+ "ar_EG": "Egypt",
375
+ "ar_SA": "Saudi Arabia",
376
+ "bg_BG": "Bulgaria",
377
+ "bn_BD": "Bangladesh",
378
+ "bn_IN": "India",
379
+ "ca_ES": "Spain",
380
+ "cs_CZ": "Czechia",
381
+ "da_DK": "Denmark",
382
+ "de_DE": "Germany",
383
+ "el_GR": "Greece",
384
+ "es_MX": "Mexico",
385
+ "et_EE": "Estonia",
386
+ "fa_IR": "Iran",
387
+ "fi_FI": "Finland",
388
+ "fil_PH": "Philippines",
389
+ "fr_CA": "Canada",
390
+ "fr_FR": "France",
391
+ "gu_IN": "India",
392
+ "he_IL": "Israel",
393
+ "hi_IN": "India",
394
+ "hr_HR": "Croatia",
395
+ "hu_HU": "Hungary",
396
+ "id_ID": "Indonesia",
397
+ "is_IS": "Iceland",
398
+ "it_IT": "Italy",
399
+ "ja_JP": "Japan",
400
+ "kn_IN": "India",
401
+ "ko_KR": "South Korea",
402
+ "lt_LT": "Lithuania",
403
+ "lv_LV": "Latvia",
404
+ "ml_IN": "India",
405
+ "mr_IN": "India",
406
+ "nl_NL": "Netherlands",
407
+ "no_NO": "Norway",
408
+ "pa_IN": "India",
409
+ "pl_PL": "Poland",
410
+ "pt_BR": "Brazil",
411
+ "pt_PT": "Portugal",
412
+ "ro_RO": "Romania",
413
+ "ru_RU": "Russia",
414
+ "sk_SK": "Slovakia",
415
+ "sl_SI": "Slovenia",
416
+ "sr_RS": "Serbia",
417
+ "sv_SE": "Sweden",
418
+ "sw_KE": "Kenya",
419
+ "sw_TZ": "Tanzania",
420
+ "ta_IN": "India",
421
+ "te_IN": "India",
422
+ "th_TH": "Thailand",
423
+ "tr_TR": "Turkey",
424
+ "uk_UA": "Ukraine",
425
+ "ur_PK": "Pakistan",
426
+ "vi_VN": "Vietnam",
427
+ "zh_CN": "China",
428
+ "zh_TW": "Taiwan",
429
+ "zu_ZA": "South Africa",
430
+ }
431
+ ```