lbourdois commited on
Commit
8b5803e
1 Parent(s): 001cb5f

Add multilingual to the language tag

Browse files

Hi! A PR to add multilingual to the language tag to improve the referencing.

Files changed (1) hide show
  1. README.md +147 -552
README.md CHANGED
@@ -12,757 +12,352 @@ language:
12
  - nl
13
  - pdc
14
  - yi
 
 
15
  tags:
16
  - translation
17
  - opus-mt-tc
18
- license: cc-by-4.0
19
  model-index:
20
  - name: opus-mt-tc-base-gmw-gmw
21
  results:
22
  - task:
23
- name: Translation afr-deu
24
  type: translation
25
- args: afr-deu
26
  dataset:
27
  name: flores101-devtest
28
  type: flores_101
29
  args: afr deu devtest
30
  metrics:
31
- - name: BLEU
32
- type: bleu
33
  value: 21.6
34
- - task:
35
- name: Translation afr-eng
36
- type: translation
37
- args: afr-eng
38
- dataset:
39
- name: flores101-devtest
40
- type: flores_101
41
- args: afr eng devtest
42
- metrics:
43
- - name: BLEU
44
- type: bleu
45
  value: 46.8
46
- - task:
47
- name: Translation deu-afr
48
- type: translation
49
- args: deu-afr
50
- dataset:
51
- name: flores101-devtest
52
- type: flores_101
53
- args: deu afr devtest
54
- metrics:
55
- - name: BLEU
56
- type: bleu
57
  value: 21.4
58
- - task:
59
- name: Translation deu-eng
60
- type: translation
61
- args: deu-eng
62
- dataset:
63
- name: flores101-devtest
64
- type: flores_101
65
- args: deu eng devtest
66
- metrics:
67
- - name: BLEU
68
- type: bleu
69
  value: 33.8
70
- - task:
71
- name: Translation eng-afr
72
- type: translation
73
- args: eng-afr
74
- dataset:
75
- name: flores101-devtest
76
- type: flores_101
77
- args: eng afr devtest
78
- metrics:
79
- - name: BLEU
80
- type: bleu
81
  value: 33.8
82
- - task:
83
- name: Translation eng-deu
84
- type: translation
85
- args: eng-deu
86
- dataset:
87
- name: flores101-devtest
88
- type: flores_101
89
- args: eng deu devtest
90
- metrics:
91
- - name: BLEU
92
- type: bleu
93
  value: 29.1
94
- - task:
95
- name: Translation eng-nld
96
- type: translation
97
- args: eng-nld
98
- dataset:
99
- name: flores101-devtest
100
- type: flores_101
101
- args: eng nld devtest
102
- metrics:
103
- - name: BLEU
104
- type: bleu
105
  value: 21.0
106
- - task:
107
- name: Translation nld-eng
108
- type: translation
109
- args: nld-eng
110
- dataset:
111
- name: flores101-devtest
112
- type: flores_101
113
- args: nld eng devtest
114
- metrics:
115
- - name: BLEU
116
- type: bleu
117
  value: 25.6
 
118
  - task:
119
- name: Translation deu-eng
120
  type: translation
121
- args: deu-eng
122
  dataset:
123
  name: multi30k_test_2016_flickr
124
  type: multi30k-2016_flickr
125
  args: deu-eng
126
  metrics:
127
- - name: BLEU
128
- type: bleu
129
  value: 32.2
130
- - task:
131
- name: Translation eng-deu
132
- type: translation
133
- args: eng-deu
134
- dataset:
135
- name: multi30k_test_2016_flickr
136
- type: multi30k-2016_flickr
137
- args: eng-deu
138
- metrics:
139
- - name: BLEU
140
- type: bleu
141
  value: 28.8
 
142
  - task:
143
- name: Translation deu-eng
144
  type: translation
145
- args: deu-eng
146
  dataset:
147
  name: multi30k_test_2017_flickr
148
  type: multi30k-2017_flickr
149
  args: deu-eng
150
  metrics:
151
- - name: BLEU
152
- type: bleu
153
  value: 32.7
154
- - task:
155
- name: Translation eng-deu
156
- type: translation
157
- args: eng-deu
158
- dataset:
159
- name: multi30k_test_2017_flickr
160
- type: multi30k-2017_flickr
161
- args: eng-deu
162
- metrics:
163
- - name: BLEU
164
- type: bleu
165
  value: 27.6
 
166
  - task:
167
- name: Translation deu-eng
168
  type: translation
169
- args: deu-eng
170
  dataset:
171
  name: multi30k_test_2017_mscoco
172
  type: multi30k-2017_mscoco
173
  args: deu-eng
174
  metrics:
175
- - name: BLEU
176
- type: bleu
177
  value: 25.5
178
- - task:
179
- name: Translation eng-deu
180
- type: translation
181
- args: eng-deu
182
- dataset:
183
- name: multi30k_test_2017_mscoco
184
- type: multi30k-2017_mscoco
185
- args: eng-deu
186
- metrics:
187
- - name: BLEU
188
- type: bleu
189
  value: 22.0
 
190
  - task:
191
- name: Translation deu-eng
192
  type: translation
193
- args: deu-eng
194
  dataset:
195
  name: multi30k_test_2018_flickr
196
  type: multi30k-2018_flickr
197
  args: deu-eng
198
  metrics:
199
- - name: BLEU
200
- type: bleu
201
  value: 30.0
202
- - task:
203
- name: Translation eng-deu
204
- type: translation
205
- args: eng-deu
206
- dataset:
207
- name: multi30k_test_2018_flickr
208
- type: multi30k-2018_flickr
209
- args: eng-deu
210
- metrics:
211
- - name: BLEU
212
- type: bleu
213
  value: 25.3
 
214
  - task:
215
- name: Translation deu-eng
216
  type: translation
217
- args: deu-eng
218
  dataset:
219
  name: news-test2008
220
  type: news-test2008
221
  args: deu-eng
222
  metrics:
223
- - name: BLEU
224
- type: bleu
225
  value: 23.8
 
226
  - task:
227
- name: Translation afr-deu
228
  type: translation
229
- args: afr-deu
230
  dataset:
231
  name: tatoeba-test-v2021-08-07
232
  type: tatoeba_mt
233
  args: afr-deu
234
  metrics:
235
- - name: BLEU
236
- type: bleu
237
  value: 48.1
238
- - task:
239
- name: Translation afr-eng
240
- type: translation
241
- args: afr-eng
242
- dataset:
243
- name: tatoeba-test-v2021-08-07
244
- type: tatoeba_mt
245
- args: afr-eng
246
- metrics:
247
- - name: BLEU
248
- type: bleu
249
  value: 58.8
250
- - task:
251
- name: Translation afr-nld
252
- type: translation
253
- args: afr-nld
254
- dataset:
255
- name: tatoeba-test-v2021-08-07
256
- type: tatoeba_mt
257
- args: afr-nld
258
- metrics:
259
- - name: BLEU
260
- type: bleu
261
  value: 54.5
262
- - task:
263
- name: Translation deu-afr
264
- type: translation
265
- args: deu-afr
266
- dataset:
267
- name: tatoeba-test-v2021-08-07
268
- type: tatoeba_mt
269
- args: deu-afr
270
- metrics:
271
- - name: BLEU
272
- type: bleu
273
  value: 52.4
274
- - task:
275
- name: Translation deu-eng
276
- type: translation
277
- args: deu-eng
278
- dataset:
279
- name: tatoeba-test-v2021-08-07
280
- type: tatoeba_mt
281
- args: deu-eng
282
- metrics:
283
- - name: BLEU
284
- type: bleu
285
  value: 42.1
286
- - task:
287
- name: Translation deu-nld
288
- type: translation
289
- args: deu-nld
290
- dataset:
291
- name: tatoeba-test-v2021-08-07
292
- type: tatoeba_mt
293
- args: deu-nld
294
- metrics:
295
- - name: BLEU
296
- type: bleu
297
  value: 48.7
298
- - task:
299
- name: Translation eng-afr
300
- type: translation
301
- args: eng-afr
302
- dataset:
303
- name: tatoeba-test-v2021-08-07
304
- type: tatoeba_mt
305
- args: eng-afr
306
- metrics:
307
- - name: BLEU
308
- type: bleu
309
  value: 56.5
310
- - task:
311
- name: Translation eng-deu
312
- type: translation
313
- args: eng-deu
314
- dataset:
315
- name: tatoeba-test-v2021-08-07
316
- type: tatoeba_mt
317
- args: eng-deu
318
- metrics:
319
- - name: BLEU
320
- type: bleu
321
  value: 35.9
322
- - task:
323
- name: Translation eng-nld
324
- type: translation
325
- args: eng-nld
326
- dataset:
327
- name: tatoeba-test-v2021-08-07
328
- type: tatoeba_mt
329
- args: eng-nld
330
- metrics:
331
- - name: BLEU
332
- type: bleu
333
  value: 48.3
334
- - task:
335
- name: Translation fry-eng
336
- type: translation
337
- args: fry-eng
338
- dataset:
339
- name: tatoeba-test-v2021-08-07
340
- type: tatoeba_mt
341
- args: fry-eng
342
- metrics:
343
- - name: BLEU
344
- type: bleu
345
  value: 32.5
346
- - task:
347
- name: Translation fry-nld
348
- type: translation
349
- args: fry-nld
350
- dataset:
351
- name: tatoeba-test-v2021-08-07
352
- type: tatoeba_mt
353
- args: fry-nld
354
- metrics:
355
- - name: BLEU
356
- type: bleu
357
  value: 43.1
358
- - task:
359
- name: Translation hrx-deu
360
- type: translation
361
- args: hrx-deu
362
- dataset:
363
- name: tatoeba-test-v2021-08-07
364
- type: tatoeba_mt
365
- args: hrx-deu
366
- metrics:
367
- - name: BLEU
368
- type: bleu
369
  value: 24.7
370
- - task:
371
- name: Translation hrx-eng
372
- type: translation
373
- args: hrx-eng
374
- dataset:
375
- name: tatoeba-test-v2021-08-07
376
- type: tatoeba_mt
377
- args: hrx-eng
378
- metrics:
379
- - name: BLEU
380
- type: bleu
381
  value: 20.4
382
- - task:
383
- name: Translation ltz-deu
384
- type: translation
385
- args: ltz-deu
386
- dataset:
387
- name: tatoeba-test-v2021-08-07
388
- type: tatoeba_mt
389
- args: ltz-deu
390
- metrics:
391
- - name: BLEU
392
- type: bleu
393
  value: 37.2
394
- - task:
395
- name: Translation ltz-eng
396
- type: translation
397
- args: ltz-eng
398
- dataset:
399
- name: tatoeba-test-v2021-08-07
400
- type: tatoeba_mt
401
- args: ltz-eng
402
- metrics:
403
- - name: BLEU
404
- type: bleu
405
  value: 32.4
406
- - task:
407
- name: Translation ltz-nld
408
- type: translation
409
- args: ltz-nld
410
- dataset:
411
- name: tatoeba-test-v2021-08-07
412
- type: tatoeba_mt
413
- args: ltz-nld
414
- metrics:
415
- - name: BLEU
416
- type: bleu
417
  value: 39.3
418
- - task:
419
- name: Translation nds-deu
420
- type: translation
421
- args: nds-deu
422
- dataset:
423
- name: tatoeba-test-v2021-08-07
424
- type: tatoeba_mt
425
- args: nds-deu
426
- metrics:
427
- - name: BLEU
428
- type: bleu
429
  value: 34.5
430
- - task:
431
- name: Translation nds-eng
432
- type: translation
433
- args: nds-eng
434
- dataset:
435
- name: tatoeba-test-v2021-08-07
436
- type: tatoeba_mt
437
- args: nds-eng
438
- metrics:
439
- - name: BLEU
440
- type: bleu
441
  value: 29.9
442
- - task:
443
- name: Translation nds-nld
444
- type: translation
445
- args: nds-nld
446
- dataset:
447
- name: tatoeba-test-v2021-08-07
448
- type: tatoeba_mt
449
- args: nds-nld
450
- metrics:
451
- - name: BLEU
452
- type: bleu
453
  value: 42.3
454
- - task:
455
- name: Translation nld-afr
456
- type: translation
457
- args: nld-afr
458
- dataset:
459
- name: tatoeba-test-v2021-08-07
460
- type: tatoeba_mt
461
- args: nld-afr
462
- metrics:
463
- - name: BLEU
464
- type: bleu
465
  value: 58.8
466
- - task:
467
- name: Translation nld-deu
468
- type: translation
469
- args: nld-deu
470
- dataset:
471
- name: tatoeba-test-v2021-08-07
472
- type: tatoeba_mt
473
- args: nld-deu
474
- metrics:
475
- - name: BLEU
476
- type: bleu
477
  value: 50.4
478
- - task:
479
- name: Translation nld-eng
480
- type: translation
481
- args: nld-eng
482
- dataset:
483
- name: tatoeba-test-v2021-08-07
484
- type: tatoeba_mt
485
- args: nld-eng
486
- metrics:
487
- - name: BLEU
488
- type: bleu
489
  value: 53.1
490
- - task:
491
- name: Translation nld-fry
492
- type: translation
493
- args: nld-fry
494
- dataset:
495
- name: tatoeba-test-v2021-08-07
496
- type: tatoeba_mt
497
- args: nld-fry
498
- metrics:
499
- - name: BLEU
500
- type: bleu
501
  value: 25.1
502
- - task:
503
- name: Translation nld-nds
504
- type: translation
505
- args: nld-nds
506
- dataset:
507
- name: tatoeba-test-v2021-08-07
508
- type: tatoeba_mt
509
- args: nld-nds
510
- metrics:
511
- - name: BLEU
512
- type: bleu
513
  value: 21.4
 
514
  - task:
515
- name: Translation deu-eng
516
  type: translation
517
- args: deu-eng
518
  dataset:
519
  name: newstest2009
520
  type: wmt-2009-news
521
  args: deu-eng
522
  metrics:
523
- - name: BLEU
524
- type: bleu
525
  value: 23.4
 
526
  - task:
527
- name: Translation deu-eng
528
  type: translation
529
- args: deu-eng
530
  dataset:
531
  name: newstest2010
532
  type: wmt-2010-news
533
  args: deu-eng
534
  metrics:
535
- - name: BLEU
536
- type: bleu
537
  value: 25.8
538
- - task:
539
- name: Translation eng-deu
540
- type: translation
541
- args: eng-deu
542
- dataset:
543
- name: newstest2010
544
- type: wmt-2010-news
545
- args: eng-deu
546
- metrics:
547
- - name: BLEU
548
- type: bleu
549
  value: 20.7
 
550
  - task:
551
- name: Translation deu-eng
552
  type: translation
553
- args: deu-eng
554
  dataset:
555
  name: newstest2011
556
  type: wmt-2011-news
557
  args: deu-eng
558
  metrics:
559
- - name: BLEU
560
- type: bleu
561
  value: 23.7
 
562
  - task:
563
- name: Translation deu-eng
564
  type: translation
565
- args: deu-eng
566
  dataset:
567
  name: newstest2012
568
  type: wmt-2012-news
569
  args: deu-eng
570
  metrics:
571
- - name: BLEU
572
- type: bleu
573
  value: 24.8
 
574
  - task:
575
- name: Translation deu-eng
576
  type: translation
577
- args: deu-eng
578
  dataset:
579
  name: newstest2013
580
  type: wmt-2013-news
581
  args: deu-eng
582
  metrics:
583
- - name: BLEU
584
- type: bleu
585
  value: 27.7
586
- - task:
587
- name: Translation eng-deu
588
- type: translation
589
- args: eng-deu
590
- dataset:
591
- name: newstest2013
592
- type: wmt-2013-news
593
- args: eng-deu
594
- metrics:
595
- - name: BLEU
596
- type: bleu
597
  value: 22.5
 
598
  - task:
599
- name: Translation deu-eng
600
  type: translation
601
- args: deu-eng
602
  dataset:
603
  name: newstest2014-deen
604
  type: wmt-2014-news
605
  args: deu-eng
606
  metrics:
607
- - name: BLEU
608
- type: bleu
609
  value: 27.3
610
- - task:
611
- name: Translation eng-deu
612
- type: translation
613
- args: eng-deu
614
- dataset:
615
- name: newstest2014-deen
616
- type: wmt-2014-news
617
- args: eng-deu
618
- metrics:
619
- - name: BLEU
620
- type: bleu
621
  value: 22.0
 
622
  - task:
623
- name: Translation deu-eng
624
  type: translation
625
- args: deu-eng
626
  dataset:
627
  name: newstest2015-deen
628
  type: wmt-2015-news
629
  args: deu-eng
630
  metrics:
631
- - name: BLEU
632
- type: bleu
633
  value: 28.6
634
- - task:
635
- name: Translation eng-deu
636
- type: translation
637
- args: eng-deu
638
- dataset:
639
- name: newstest2015-ende
640
- type: wmt-2015-news
641
- args: eng-deu
642
- metrics:
643
- - name: BLEU
644
- type: bleu
645
  value: 25.7
 
646
  - task:
647
- name: Translation deu-eng
648
  type: translation
649
- args: deu-eng
650
  dataset:
651
  name: newstest2016-deen
652
  type: wmt-2016-news
653
  args: deu-eng
654
  metrics:
655
- - name: BLEU
656
- type: bleu
657
  value: 33.3
658
- - task:
659
- name: Translation eng-deu
660
- type: translation
661
- args: eng-deu
662
- dataset:
663
- name: newstest2016-ende
664
- type: wmt-2016-news
665
- args: eng-deu
666
- metrics:
667
- - name: BLEU
668
- type: bleu
669
  value: 30.0
 
670
  - task:
671
- name: Translation deu-eng
672
  type: translation
673
- args: deu-eng
674
  dataset:
675
  name: newstest2017-deen
676
  type: wmt-2017-news
677
  args: deu-eng
678
  metrics:
679
- - name: BLEU
680
- type: bleu
681
  value: 29.5
682
- - task:
683
- name: Translation eng-deu
684
- type: translation
685
- args: eng-deu
686
- dataset:
687
- name: newstest2017-ende
688
- type: wmt-2017-news
689
- args: eng-deu
690
- metrics:
691
- - name: BLEU
692
- type: bleu
693
  value: 24.1
 
694
  - task:
695
- name: Translation deu-eng
696
  type: translation
697
- args: deu-eng
698
  dataset:
699
  name: newstest2018-deen
700
  type: wmt-2018-news
701
  args: deu-eng
702
  metrics:
703
- - name: BLEU
704
- type: bleu
705
  value: 36.1
706
- - task:
707
- name: Translation eng-deu
708
- type: translation
709
- args: eng-deu
710
- dataset:
711
- name: newstest2018-ende
712
- type: wmt-2018-news
713
- args: eng-deu
714
- metrics:
715
- - name: BLEU
716
- type: bleu
717
  value: 35.4
 
718
  - task:
719
- name: Translation deu-eng
720
  type: translation
721
- args: deu-eng
722
  dataset:
723
  name: newstest2019-deen
724
  type: wmt-2019-news
725
  args: deu-eng
726
  metrics:
727
- - name: BLEU
728
- type: bleu
729
  value: 32.3
730
- - task:
731
- name: Translation eng-deu
732
- type: translation
733
- args: eng-deu
734
- dataset:
735
- name: newstest2019-ende
736
- type: wmt-2019-news
737
- args: eng-deu
738
- metrics:
739
- - name: BLEU
740
- type: bleu
741
  value: 31.2
 
742
  - task:
743
- name: Translation deu-eng
744
  type: translation
745
- args: deu-eng
746
  dataset:
747
  name: newstest2020-deen
748
  type: wmt-2020-news
749
  args: deu-eng
750
  metrics:
751
- - name: BLEU
752
- type: bleu
753
  value: 32.0
754
- - task:
755
- name: Translation eng-deu
756
- type: translation
757
- args: eng-deu
758
- dataset:
759
- name: newstest2020-ende
760
- type: wmt-2020-news
761
- args: eng-deu
762
- metrics:
763
- - name: BLEU
764
- type: bleu
765
  value: 23.9
 
766
  ---
767
  # opus-mt-tc-base-gmw-gmw
768
 
@@ -770,7 +365,7 @@ Neural machine translation model for translating from West Germanic languages (g
770
 
771
  This model is part of the [OPUS-MT project](https://github.com/Helsinki-NLP/Opus-MT), an effort to make neural machine translation models widely available and accessible for many languages in the world. All models are originally trained using the amazing framework of [Marian NMT](https://marian-nmt.github.io/), an efficient NMT implementation written in pure C++. The models have been converted to pyTorch using the transformers library by huggingface. Training data is taken from [OPUS](https://opus.nlpl.eu/) and training pipelines use the procedures of [OPUS-MT-train](https://github.com/Helsinki-NLP/Opus-MT-train).
772
 
773
- * Publications: [OPUS-MT Building open translation services for the World](https://aclanthology.org/2020.eamt-1.61/) and [The Tatoeba Translation Challenge Realistic Data Sets for Low Resource and Multilingual MT](https://aclanthology.org/2020.wmt-1.139/) (Please, cite if you use this model.)
774
 
775
  ```
776
  @inproceedings{tiedemann-thottingal-2020-opus,
@@ -941,7 +536,7 @@ print(pipe(>>nld<< You need help.))
941
 
942
  ## Acknowledgements
943
 
944
- The work is supported by the [European Language Grid](https://www.european-language-grid.eu/) as [pilot project 2866](https://live.european-language-grid.eu/catalogue/#/resource/projects/2866), by the [FoTran project](https://www.helsinki.fi/en/researchgroups/natural-language-understanding-with-cross-lingual-grounding), funded by the European Research Council (ERC) under the European Unions Horizon 2020 research and innovation programme (grant agreement No 771113), and the [MeMAD project](https://memad.eu/), funded by the European Unions Horizon 2020 Research and Innovation Programme under grant agreement No 780069. We are also grateful for the generous computational resources and IT infrastructure provided by [CSC -- IT Center for Science](https://www.csc.fi/), Finland.
945
 
946
  ## Model conversion info
947
 
 
12
  - nl
13
  - pdc
14
  - yi
15
+ - multilingual
16
+ license: cc-by-4.0
17
  tags:
18
  - translation
19
  - opus-mt-tc
 
20
  model-index:
21
  - name: opus-mt-tc-base-gmw-gmw
22
  results:
23
  - task:
 
24
  type: translation
25
+ name: Translation afr-deu
26
  dataset:
27
  name: flores101-devtest
28
  type: flores_101
29
  args: afr deu devtest
30
  metrics:
31
+ - type: bleu
 
32
  value: 21.6
33
+ name: BLEU
34
+ - type: bleu
 
 
 
 
 
 
 
 
 
35
  value: 46.8
36
+ name: BLEU
37
+ - type: bleu
 
 
 
 
 
 
 
 
 
38
  value: 21.4
39
+ name: BLEU
40
+ - type: bleu
 
 
 
 
 
 
 
 
 
41
  value: 33.8
42
+ name: BLEU
43
+ - type: bleu
 
 
 
 
 
 
 
 
 
44
  value: 33.8
45
+ name: BLEU
46
+ - type: bleu
 
 
 
 
 
 
 
 
 
47
  value: 29.1
48
+ name: BLEU
49
+ - type: bleu
 
 
 
 
 
 
 
 
 
50
  value: 21.0
51
+ name: BLEU
52
+ - type: bleu
 
 
 
 
 
 
 
 
 
53
  value: 25.6
54
+ name: BLEU
55
  - task:
 
56
  type: translation
57
+ name: Translation deu-eng
58
  dataset:
59
  name: multi30k_test_2016_flickr
60
  type: multi30k-2016_flickr
61
  args: deu-eng
62
  metrics:
63
+ - type: bleu
 
64
  value: 32.2
65
+ name: BLEU
66
+ - type: bleu
 
 
 
 
 
 
 
 
 
67
  value: 28.8
68
+ name: BLEU
69
  - task:
 
70
  type: translation
71
+ name: Translation deu-eng
72
  dataset:
73
  name: multi30k_test_2017_flickr
74
  type: multi30k-2017_flickr
75
  args: deu-eng
76
  metrics:
77
+ - type: bleu
 
78
  value: 32.7
79
+ name: BLEU
80
+ - type: bleu
 
 
 
 
 
 
 
 
 
81
  value: 27.6
82
+ name: BLEU
83
  - task:
 
84
  type: translation
85
+ name: Translation deu-eng
86
  dataset:
87
  name: multi30k_test_2017_mscoco
88
  type: multi30k-2017_mscoco
89
  args: deu-eng
90
  metrics:
91
+ - type: bleu
 
92
  value: 25.5
93
+ name: BLEU
94
+ - type: bleu
 
 
 
 
 
 
 
 
 
95
  value: 22.0
96
+ name: BLEU
97
  - task:
 
98
  type: translation
99
+ name: Translation deu-eng
100
  dataset:
101
  name: multi30k_test_2018_flickr
102
  type: multi30k-2018_flickr
103
  args: deu-eng
104
  metrics:
105
+ - type: bleu
 
106
  value: 30.0
107
+ name: BLEU
108
+ - type: bleu
 
 
 
 
 
 
 
 
 
109
  value: 25.3
110
+ name: BLEU
111
  - task:
 
112
  type: translation
113
+ name: Translation deu-eng
114
  dataset:
115
  name: news-test2008
116
  type: news-test2008
117
  args: deu-eng
118
  metrics:
119
+ - type: bleu
 
120
  value: 23.8
121
+ name: BLEU
122
  - task:
 
123
  type: translation
124
+ name: Translation afr-deu
125
  dataset:
126
  name: tatoeba-test-v2021-08-07
127
  type: tatoeba_mt
128
  args: afr-deu
129
  metrics:
130
+ - type: bleu
 
131
  value: 48.1
132
+ name: BLEU
133
+ - type: bleu
 
 
 
 
 
 
 
 
 
134
  value: 58.8
135
+ name: BLEU
136
+ - type: bleu
 
 
 
 
 
 
 
 
 
137
  value: 54.5
138
+ name: BLEU
139
+ - type: bleu
 
 
 
 
 
 
 
 
 
140
  value: 52.4
141
+ name: BLEU
142
+ - type: bleu
 
 
 
 
 
 
 
 
 
143
  value: 42.1
144
+ name: BLEU
145
+ - type: bleu
 
 
 
 
 
 
 
 
 
146
  value: 48.7
147
+ name: BLEU
148
+ - type: bleu
 
 
 
 
 
 
 
 
 
149
  value: 56.5
150
+ name: BLEU
151
+ - type: bleu
 
 
 
 
 
 
 
 
 
152
  value: 35.9
153
+ name: BLEU
154
+ - type: bleu
 
 
 
 
 
 
 
 
 
155
  value: 48.3
156
+ name: BLEU
157
+ - type: bleu
 
 
 
 
 
 
 
 
 
158
  value: 32.5
159
+ name: BLEU
160
+ - type: bleu
 
 
 
 
 
 
 
 
 
161
  value: 43.1
162
+ name: BLEU
163
+ - type: bleu
 
 
 
 
 
 
 
 
 
164
  value: 24.7
165
+ name: BLEU
166
+ - type: bleu
 
 
 
 
 
 
 
 
 
167
  value: 20.4
168
+ name: BLEU
169
+ - type: bleu
 
 
 
 
 
 
 
 
 
170
  value: 37.2
171
+ name: BLEU
172
+ - type: bleu
 
 
 
 
 
 
 
 
 
173
  value: 32.4
174
+ name: BLEU
175
+ - type: bleu
 
 
 
 
 
 
 
 
 
176
  value: 39.3
177
+ name: BLEU
178
+ - type: bleu
 
 
 
 
 
 
 
 
 
179
  value: 34.5
180
+ name: BLEU
181
+ - type: bleu
 
 
 
 
 
 
 
 
 
182
  value: 29.9
183
+ name: BLEU
184
+ - type: bleu
 
 
 
 
 
 
 
 
 
185
  value: 42.3
186
+ name: BLEU
187
+ - type: bleu
 
 
 
 
 
 
 
 
 
188
  value: 58.8
189
+ name: BLEU
190
+ - type: bleu
 
 
 
 
 
 
 
 
 
191
  value: 50.4
192
+ name: BLEU
193
+ - type: bleu
 
 
 
 
 
 
 
 
 
194
  value: 53.1
195
+ name: BLEU
196
+ - type: bleu
 
 
 
 
 
 
 
 
 
197
  value: 25.1
198
+ name: BLEU
199
+ - type: bleu
 
 
 
 
 
 
 
 
 
200
  value: 21.4
201
+ name: BLEU
202
  - task:
 
203
  type: translation
204
+ name: Translation deu-eng
205
  dataset:
206
  name: newstest2009
207
  type: wmt-2009-news
208
  args: deu-eng
209
  metrics:
210
+ - type: bleu
 
211
  value: 23.4
212
+ name: BLEU
213
  - task:
 
214
  type: translation
215
+ name: Translation deu-eng
216
  dataset:
217
  name: newstest2010
218
  type: wmt-2010-news
219
  args: deu-eng
220
  metrics:
221
+ - type: bleu
 
222
  value: 25.8
223
+ name: BLEU
224
+ - type: bleu
 
 
 
 
 
 
 
 
 
225
  value: 20.7
226
+ name: BLEU
227
  - task:
 
228
  type: translation
229
+ name: Translation deu-eng
230
  dataset:
231
  name: newstest2011
232
  type: wmt-2011-news
233
  args: deu-eng
234
  metrics:
235
+ - type: bleu
 
236
  value: 23.7
237
+ name: BLEU
238
  - task:
 
239
  type: translation
240
+ name: Translation deu-eng
241
  dataset:
242
  name: newstest2012
243
  type: wmt-2012-news
244
  args: deu-eng
245
  metrics:
246
+ - type: bleu
 
247
  value: 24.8
248
+ name: BLEU
249
  - task:
 
250
  type: translation
251
+ name: Translation deu-eng
252
  dataset:
253
  name: newstest2013
254
  type: wmt-2013-news
255
  args: deu-eng
256
  metrics:
257
+ - type: bleu
 
258
  value: 27.7
259
+ name: BLEU
260
+ - type: bleu
 
 
 
 
 
 
 
 
 
261
  value: 22.5
262
+ name: BLEU
263
  - task:
 
264
  type: translation
265
+ name: Translation deu-eng
266
  dataset:
267
  name: newstest2014-deen
268
  type: wmt-2014-news
269
  args: deu-eng
270
  metrics:
271
+ - type: bleu
 
272
  value: 27.3
273
+ name: BLEU
274
+ - type: bleu
 
 
 
 
 
 
 
 
 
275
  value: 22.0
276
+ name: BLEU
277
  - task:
 
278
  type: translation
279
+ name: Translation deu-eng
280
  dataset:
281
  name: newstest2015-deen
282
  type: wmt-2015-news
283
  args: deu-eng
284
  metrics:
285
+ - type: bleu
 
286
  value: 28.6
287
+ name: BLEU
288
+ - type: bleu
 
 
 
 
 
 
 
 
 
289
  value: 25.7
290
+ name: BLEU
291
  - task:
 
292
  type: translation
293
+ name: Translation deu-eng
294
  dataset:
295
  name: newstest2016-deen
296
  type: wmt-2016-news
297
  args: deu-eng
298
  metrics:
299
+ - type: bleu
 
300
  value: 33.3
301
+ name: BLEU
302
+ - type: bleu
 
 
 
 
 
 
 
 
 
303
  value: 30.0
304
+ name: BLEU
305
  - task:
 
306
  type: translation
307
+ name: Translation deu-eng
308
  dataset:
309
  name: newstest2017-deen
310
  type: wmt-2017-news
311
  args: deu-eng
312
  metrics:
313
+ - type: bleu
 
314
  value: 29.5
315
+ name: BLEU
316
+ - type: bleu
 
 
 
 
 
 
 
 
 
317
  value: 24.1
318
+ name: BLEU
319
  - task:
 
320
  type: translation
321
+ name: Translation deu-eng
322
  dataset:
323
  name: newstest2018-deen
324
  type: wmt-2018-news
325
  args: deu-eng
326
  metrics:
327
+ - type: bleu
 
328
  value: 36.1
329
+ name: BLEU
330
+ - type: bleu
 
 
 
 
 
 
 
 
 
331
  value: 35.4
332
+ name: BLEU
333
  - task:
 
334
  type: translation
335
+ name: Translation deu-eng
336
  dataset:
337
  name: newstest2019-deen
338
  type: wmt-2019-news
339
  args: deu-eng
340
  metrics:
341
+ - type: bleu
 
342
  value: 32.3
343
+ name: BLEU
344
+ - type: bleu
 
 
 
 
 
 
 
 
 
345
  value: 31.2
346
+ name: BLEU
347
  - task:
 
348
  type: translation
349
+ name: Translation deu-eng
350
  dataset:
351
  name: newstest2020-deen
352
  type: wmt-2020-news
353
  args: deu-eng
354
  metrics:
355
+ - type: bleu
 
356
  value: 32.0
357
+ name: BLEU
358
+ - type: bleu
 
 
 
 
 
 
 
 
 
359
  value: 23.9
360
+ name: BLEU
361
  ---
362
  # opus-mt-tc-base-gmw-gmw
363
 
 
365
 
366
  This model is part of the [OPUS-MT project](https://github.com/Helsinki-NLP/Opus-MT), an effort to make neural machine translation models widely available and accessible for many languages in the world. All models are originally trained using the amazing framework of [Marian NMT](https://marian-nmt.github.io/), an efficient NMT implementation written in pure C++. The models have been converted to pyTorch using the transformers library by huggingface. Training data is taken from [OPUS](https://opus.nlpl.eu/) and training pipelines use the procedures of [OPUS-MT-train](https://github.com/Helsinki-NLP/Opus-MT-train).
367
 
368
+ * Publications: [OPUS-MT Building open translation services for the World](https://aclanthology.org/2020.eamt-1.61/) and [The Tatoeba Translation Challenge Realistic Data Sets for Low Resource and Multilingual MT](https://aclanthology.org/2020.wmt-1.139/) (Please, cite if you use this model.)
369
 
370
  ```
371
  @inproceedings{tiedemann-thottingal-2020-opus,
 
536
 
537
  ## Acknowledgements
538
 
539
+ The work is supported by the [European Language Grid](https://www.european-language-grid.eu/) as [pilot project 2866](https://live.european-language-grid.eu/catalogue/#/resource/projects/2866), by the [FoTran project](https://www.helsinki.fi/en/researchgroups/natural-language-understanding-with-cross-lingual-grounding), funded by the European Research Council (ERC) under the European Unions Horizon 2020 research and innovation programme (grant agreement No 771113), and the [MeMAD project](https://memad.eu/), funded by the European Unions Horizon 2020 Research and Innovation Programme under grant agreement No 780069. We are also grateful for the generous computational resources and IT infrastructure provided by [CSC -- IT Center for Science](https://www.csc.fi/), Finland.
540
 
541
  ## Model conversion info
542