File size: 15,719 Bytes
579fe1b
 
a5fbed6
579fe1b
 
 
a5fbed6
579fe1b
 
 
 
a5fbed6
579fe1b
7fc2adf
 
579fe1b
 
 
 
 
 
 
 
7fc2adf
579fe1b
 
 
 
 
7fc2adf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
579fe1b
 
 
 
 
 
 
 
7fc2adf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
579fe1b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a5fbed6
 
 
579fe1b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7fc2adf
 
579fe1b
 
 
 
 
 
 
 
 
 
 
 
7fc2adf
579fe1b
 
 
 
 
 
 
7fc2adf
579fe1b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7fc2adf
579fe1b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7fc2adf
579fe1b
 
 
 
 
a5fbed6
579fe1b
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
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
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
---
language:
- ca
- da
- es
- fr
- gl
- is
- it
- nb
- pt
- ro
- sv
- multilingual
license: cc-by-4.0
tags:
- translation
- opus-mt-tc
model-index:
- name: opus-mt-tc-big-gmq-itc
  results:
  - task:
      type: translation
      name: Translation dan-cat
    dataset:
      name: flores101-devtest
      type: flores_101
      args: dan cat devtest
    metrics:
    - type: bleu
      value: 33.4
      name: BLEU
    - type: chrf
      value: 0.59224
      name: chr-F
    - type: bleu
      value: 38.3
      name: BLEU
    - type: chrf
      value: 0.63387
      name: chr-F
    - type: bleu
      value: 26.4
      name: BLEU
    - type: chrf
      value: 0.54446
      name: chr-F
    - type: bleu
      value: 25.7
      name: BLEU
    - type: chrf
      value: 0.55237
      name: chr-F
    - type: bleu
      value: 36.9
      name: BLEU
    - type: chrf
      value: 0.62233
      name: chr-F
    - type: bleu
      value: 31.8
      name: BLEU
    - type: chrf
      value: 0.58235
      name: chr-F
    - type: bleu
      value: 24.3
      name: BLEU
    - type: chrf
      value: 0.52453
      name: chr-F
    - type: bleu
      value: 22.7
      name: BLEU
    - type: chrf
      value: 0.4893
      name: chr-F
    - type: bleu
      value: 26.2
      name: BLEU
    - type: chrf
      value: 0.52704
      name: chr-F
    - type: bleu
      value: 18.0
      name: BLEU
    - type: chrf
      value: 0.45387
      name: chr-F
    - type: bleu
      value: 18.6
      name: BLEU
    - type: chrf
      value: 0.47303
      name: chr-F
    - type: bleu
      value: 24.9
      name: BLEU
    - type: chrf
      value: 0.51381
      name: chr-F
    - type: bleu
      value: 21.6
      name: BLEU
    - type: chrf
      value: 0.48224
      name: chr-F
    - type: bleu
      value: 18.1
      name: BLEU
    - type: chrf
      value: 0.45786
      name: chr-F
    - type: bleu
      value: 28.9
      name: BLEU
    - type: chrf
      value: 0.55984
      name: chr-F
    - type: bleu
      value: 33.8
      name: BLEU
    - type: chrf
      value: 0.60102
      name: chr-F
    - type: bleu
      value: 23.4
      name: BLEU
    - type: chrf
      value: 0.52145
      name: chr-F
    - type: bleu
      value: 22.2
      name: BLEU
    - type: chrf
      value: 0.52619
      name: chr-F
    - type: bleu
      value: 32.2
      name: BLEU
    - type: chrf
      value: 0.58836
      name: chr-F
    - type: bleu
      value: 27.6
      name: BLEU
    - type: chrf
      value: 0.54845
      name: chr-F
    - type: bleu
      value: 21.8
      name: BLEU
    - type: chrf
      value: 0.50661
      name: chr-F
    - type: bleu
      value: 32.4
      name: BLEU
    - type: chrf
      value: 0.58542
      name: chr-F
    - type: bleu
      value: 39.3
      name: BLEU
    - type: chrf
      value: 0.63688
      name: chr-F
    - type: bleu
      value: 26.0
      name: BLEU
    - type: chrf
      value: 0.53989
      name: chr-F
    - type: bleu
      value: 25.9
      name: BLEU
    - type: chrf
      value: 0.55232
      name: chr-F
    - type: bleu
      value: 36.5
      name: BLEU
    - type: chrf
      value: 0.61882
      name: chr-F
    - type: bleu
      value: 31.0
      name: BLEU
    - type: chrf
      value: 0.57419
      name: chr-F
    - type: bleu
      value: 23.8
      name: BLEU
    - type: chrf
      value: 0.52175
      name: chr-F
  - task:
      type: translation
      name: Translation dan-fra
    dataset:
      name: tatoeba-test-v2021-08-07
      type: tatoeba_mt
      args: dan-fra
    metrics:
    - type: bleu
      value: 63.8
      name: BLEU
    - type: chrf
      value: 0.76671
      name: chr-F
    - type: bleu
      value: 56.2
      name: BLEU
    - type: chrf
      value: 0.74658
      name: chr-F
    - type: bleu
      value: 57.8
      name: BLEU
    - type: chrf
      value: 0.74944
      name: chr-F
    - type: bleu
      value: 54.8
      name: BLEU
    - type: chrf
      value: 0.72328
      name: chr-F
    - type: bleu
      value: 51.0
      name: BLEU
    - type: chrf
      value: 0.69354
      name: chr-F
    - type: bleu
      value: 49.2
      name: BLEU
    - type: chrf
      value: 0.66008
      name: chr-F
    - type: bleu
      value: 54.4
      name: BLEU
    - type: chrf
      value: 0.70854
      name: chr-F
    - type: bleu
      value: 55.9
      name: BLEU
    - type: chrf
      value: 0.73672
      name: chr-F
    - type: bleu
      value: 59.2
      name: BLEU
    - type: chrf
      value: 0.73014
      name: chr-F
    - type: bleu
      value: 56.6
      name: BLEU
    - type: chrf
      value: 0.73211
      name: chr-F
    - type: bleu
      value: 48.7
      name: BLEU
    - type: chrf
      value: 0.68146
      name: chr-F
    - type: bleu
      value: 55.3
      name: BLEU
    - type: chrf
      value: 0.71373
      name: chr-F
---
# opus-mt-tc-big-gmq-itc

## Table of Contents
- [Model Details](#model-details)
- [Uses](#uses)
- [Risks, Limitations and Biases](#risks-limitations-and-biases)
- [How to Get Started With the Model](#how-to-get-started-with-the-model)
- [Training](#training)
- [Evaluation](#evaluation)
- [Citation Information](#citation-information)
- [Acknowledgements](#acknowledgements)

## Model Details

Neural machine translation model for translating from North Germanic languages (gmq) to Italic languages (itc).

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).
**Model Description:**
- **Developed by:** Language Technology Research Group at the University of Helsinki
- **Model Type:** Translation (transformer-big)
- **Release**: 2022-08-09
- **License:** CC-BY-4.0
- **Language(s):**  
  - Source Language(s): dan isl nno nob nor swe
  - Target Language(s): cat fra glg ita lat por ron spa
  - Language Pair(s): dan-cat dan-fra dan-glg dan-ita dan-por dan-ron dan-spa isl-cat isl-fra isl-ita isl-por isl-ron isl-spa nob-cat nob-fra nob-glg nob-ita nob-por nob-ron nob-spa swe-cat swe-fra swe-glg swe-ita swe-por swe-ron swe-spa
  - Valid Target Language Labels: >>acf<< >>aoa<< >>arg<< >>ast<< >>cat<< >>cbk<< >>ccd<< >>cks<< >>cos<< >>cri<< >>crs<< >>dlm<< >>drc<< >>egl<< >>ext<< >>fab<< >>fax<< >>fra<< >>frc<< >>frm<< >>fro<< >>frp<< >>fur<< >>gcf<< >>gcr<< >>glg<< >>hat<< >>idb<< >>ist<< >>ita<< >>itk<< >>kea<< >>kmv<< >>lad<< >>lad_Latn<< >>lat<< >>lat_Latn<< >>lij<< >>lld<< >>lmo<< >>lou<< >>mcm<< >>mfe<< >>mol<< >>mwl<< >>mxi<< >>mzs<< >>nap<< >>nrf<< >>oci<< >>osc<< >>osp<< >>osp_Latn<< >>pap<< >>pcd<< >>pln<< >>pms<< >>pob<< >>por<< >>pov<< >>pre<< >>pro<< >>qbb<< >>qhr<< >>rcf<< >>rgn<< >>roh<< >>ron<< >>ruo<< >>rup<< >>ruq<< >>scf<< >>scn<< >>sdc<< >>sdn<< >>spa<< >>spq<< >>spx<< >>src<< >>srd<< >>sro<< >>tmg<< >>tvy<< >>vec<< >>vkp<< >>wln<< >>xfa<< >>xum<<
- **Original Model**: [opusTCv20210807_transformer-big_2022-08-09.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/gmq-itc/opusTCv20210807_transformer-big_2022-08-09.zip)
- **Resources for more information:**
  - [OPUS-MT-train GitHub Repo](https://github.com/Helsinki-NLP/OPUS-MT-train)
  - More information about released models for this language pair: [OPUS-MT gmq-itc README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/gmq-itc/README.md)
  - [More information about MarianNMT models in the transformers library](https://huggingface.co/docs/transformers/model_doc/marian)
  - [Tatoeba Translation Challenge](https://github.com/Helsinki-NLP/Tatoeba-Challenge/

This is a multilingual translation model with multiple target languages. A sentence initial language token is required in the form of `>>id<<` (id = valid target language ID), e.g. `>>fra<<`

## Uses

This model can be used for translation and text-to-text generation.

## Risks, Limitations and Biases

**CONTENT WARNING: Readers should be aware that the model is trained on various public data sets that may contain content that is disturbing, offensive, and can propagate historical and current stereotypes.**

Significant research has explored bias and fairness issues with language models (see, e.g., [Sheng et al. (2021)](https://aclanthology.org/2021.acl-long.330.pdf) and [Bender et al. (2021)](https://dl.acm.org/doi/pdf/10.1145/3442188.3445922)).

## How to Get Started With the Model

A short example code:

```python
from transformers import MarianMTModel, MarianTokenizer

src_text = [
    ">>spa<< Jag �r inte religi�s.",
    ">>por<< Livet er for kort til � l�re seg tysk."
]

model_name = "pytorch-models/opus-mt-tc-big-gmq-itc"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))

for t in translated:
    print( tokenizer.decode(t, skip_special_tokens=True) )

# expected output:
#     No soy religioso.
#     A vida � muito curta para aprender alem�o.
```

You can also use OPUS-MT models with the transformers pipelines, for example:

```python
from transformers import pipeline
pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-gmq-itc")
print(pipe(">>spa<< Jag �r inte religi�s."))

# expected output: No soy religioso.
```

## Training

- **Data**: opusTCv20210807 ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
- **Pre-processing**: SentencePiece (spm32k,spm32k)
- **Model Type:**  transformer-big
- **Original MarianNMT Model**: [opusTCv20210807_transformer-big_2022-08-09.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/gmq-itc/opusTCv20210807_transformer-big_2022-08-09.zip)
- **Training Scripts**: [GitHub Repo](https://github.com/Helsinki-NLP/OPUS-MT-train)

## Evaluation

* test set translations: [opusTCv20210807_transformer-big_2022-08-09.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/gmq-itc/opusTCv20210807_transformer-big_2022-08-09.test.txt)
* test set scores: [opusTCv20210807_transformer-big_2022-08-09.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/gmq-itc/opusTCv20210807_transformer-big_2022-08-09.eval.txt)
* benchmark results: [benchmark_results.txt](benchmark_results.txt)
* benchmark output: [benchmark_translations.zip](benchmark_translations.zip)

| langpair | testset | chr-F | BLEU  | #sent | #words |
|----------|---------|-------|-------|-------|--------|
| dan-fra | tatoeba-test-v2021-08-07 | 0.76671 | 63.8 | 1731 | 11882 |
| dan-ita | tatoeba-test-v2021-08-07 | 0.74658 | 56.2 | 284 | 2226 |
| dan-por | tatoeba-test-v2021-08-07 | 0.74944 | 57.8 | 873 | 5360 |
| dan-spa | tatoeba-test-v2021-08-07 | 0.72328 | 54.8 | 5000 | 35528 |
| isl-ita | tatoeba-test-v2021-08-07 | 0.69354 | 51.0 | 236 | 1450 |
| isl-spa | tatoeba-test-v2021-08-07 | 0.66008 | 49.2 | 238 | 1229 |
| nob-fra | tatoeba-test-v2021-08-07 | 0.70854 | 54.4 | 323 | 2269 |
| nob-spa | tatoeba-test-v2021-08-07 | 0.73672 | 55.9 | 885 | 6866 |
| swe-fra | tatoeba-test-v2021-08-07 | 0.73014 | 59.2 | 1407 | 9580 |
| swe-ita | tatoeba-test-v2021-08-07 | 0.73211 | 56.6 | 715 | 4711 |
| swe-por | tatoeba-test-v2021-08-07 | 0.68146 | 48.7 | 320 | 2032 |
| swe-spa | tatoeba-test-v2021-08-07 | 0.71373 | 55.3 | 1351 | 8235 |
| dan-cat | flores101-devtest | 0.59224 | 33.4 | 1012 | 27304 |
| dan-fra | flores101-devtest | 0.63387 | 38.3 | 1012 | 28343 |
| dan-glg | flores101-devtest | 0.54446 | 26.4 | 1012 | 26582 |
| dan-ita | flores101-devtest | 0.55237 | 25.7 | 1012 | 27306 |
| dan-por | flores101-devtest | 0.62233 | 36.9 | 1012 | 26519 |
| dan-ron | flores101-devtest | 0.58235 | 31.8 | 1012 | 26799 |
| dan-spa | flores101-devtest | 0.52453 | 24.3 | 1012 | 29199 |
| isl-cat | flores101-devtest | 0.48930 | 22.7 | 1012 | 27304 |
| isl-fra | flores101-devtest | 0.52704 | 26.2 | 1012 | 28343 |
| isl-glg | flores101-devtest | 0.45387 | 18.0 | 1012 | 26582 |
| isl-ita | flores101-devtest | 0.47303 | 18.6 | 1012 | 27306 |
| isl-por | flores101-devtest | 0.51381 | 24.9 | 1012 | 26519 |
| isl-ron | flores101-devtest | 0.48224 | 21.6 | 1012 | 26799 |
| isl-spa | flores101-devtest | 0.45786 | 18.1 | 1012 | 29199 |
| nob-cat | flores101-devtest | 0.55984 | 28.9 | 1012 | 27304 |
| nob-fra | flores101-devtest | 0.60102 | 33.8 | 1012 | 28343 |
| nob-glg | flores101-devtest | 0.52145 | 23.4 | 1012 | 26582 |
| nob-ita | flores101-devtest | 0.52619 | 22.2 | 1012 | 27306 |
| nob-por | flores101-devtest | 0.58836 | 32.2 | 1012 | 26519 |
| nob-ron | flores101-devtest | 0.54845 | 27.6 | 1012 | 26799 |
| nob-spa | flores101-devtest | 0.50661 | 21.8 | 1012 | 29199 |
| swe-cat | flores101-devtest | 0.58542 | 32.4 | 1012 | 27304 |
| swe-fra | flores101-devtest | 0.63688 | 39.3 | 1012 | 28343 |
| swe-glg | flores101-devtest | 0.53989 | 26.0 | 1012 | 26582 |
| swe-ita | flores101-devtest | 0.55232 | 25.9 | 1012 | 27306 |
| swe-por | flores101-devtest | 0.61882 | 36.5 | 1012 | 26519 |
| swe-ron | flores101-devtest | 0.57419 | 31.0 | 1012 | 26799 |
| swe-spa | flores101-devtest | 0.52175 | 23.8 | 1012 | 29199 |

## Citation Information

* 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.)

```
@inproceedings{tiedemann-thottingal-2020-opus,
    title = "{OPUS}-{MT} {--} Building open translation services for the World",
    author = {Tiedemann, J{\"o}rg  and Thottingal, Santhosh},
    booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
    month = nov,
    year = "2020",
    address = "Lisboa, Portugal",
    publisher = "European Association for Machine Translation",
    url = "https://aclanthology.org/2020.eamt-1.61",
    pages = "479--480",
}

@inproceedings{tiedemann-2020-tatoeba,
    title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
    author = {Tiedemann, J{\"o}rg},
    booktitle = "Proceedings of the Fifth Conference on Machine Translation",
    month = nov,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.wmt-1.139",
    pages = "1174--1182",
}
```

## Acknowledgements

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 Union�s Horizon 2020 research and innovation programme (grant agreement No 771113), and the [MeMAD project](https://memad.eu/), funded by the European Union�s 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.

## Model conversion info

* transformers version: 4.16.2
* OPUS-MT git hash: 8b9f0b0
* port time: Sat Aug 13 00:00:00 EEST 2022
* port machine: LM0-400-22516.local