File size: 9,514 Bytes
876a69c
7ca3cf7
876a69c
310609f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
43b643b
310609f
 
 
 
 
 
 
 
 
 
 
 
 
8a88937
310609f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7b11e16
310609f
 
 
 
 
43b643b
310609f
7b11e16
 
 
 
 
 
 
 
310609f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15954f2
310609f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license:  apache-2.0
---
## PlanTL Project's Spanish-Catalan  machine translation model

## Table of Contents
- [Model Description](#model-description)
- [Intended Uses and Limitations](#intended-use)
- [How to Use](#how-to-use)
- [Training](#training)
  - [Training data](#training-data)
  - [Training procedure](#training-procedure)
    - [Data Preparation](#data-preparation)
    - [Tokenization](#tokenization)
    - [Hyperparameters](#hyperparameters)
- [Evaluation](#evaluation)
   - [Variable and Metrics](#variable-and-metrics)
   - [Evaluation Results](#evaluation-results)
- [Additional Information](#additional-information)
  - [Author](#author)
  - [Contact Information](#contact-information)
  - [Copyright](#copyright)
  - [Licensing Information](#licensing-information)
  - [Funding](#funding)
  - [Disclaimer](#disclaimer)
  
## Model description

This model was trained from scratch using the [Fairseq toolkit](https://fairseq.readthedocs.io/en/latest/) on a combination of Spanish-Catalan datasets, up to 92 million sentences. Additionally, the model is evaluated on several public datasecomprising 5 different domains (general, adminstrative, technology, biomedical, and news).  

## Intended uses and limitations

You can use this model for machine translation from Spanish to Catalan. 

## How to use

### Usage
Required libraries:

```bash
pip install ctranslate2 pyonmttok
```

Translate a sentence using python 
```python
import ctranslate2
import pyonmttok
from huggingface_hub import snapshot_download
model_dir = snapshot_download(repo_id="PlanTL-GOB-ES/mt-plantl-es-ca", revision="main")

tokenizer=pyonmttok.Tokenizer(mode="none", sp_model_path = model_dir + "/spm.model")
tokenized=tokenizer.tokenize("Bienvenido al Proyecto PlanTL!")

translator = ctranslate2.Translator(model_dir)
translated = translator.translate_batch([tokenized[0]])
print(tokenizer.detokenize(translated[0][0]['tokens']))
```

## Training

### Training data

The model was trained on a combination of the following datasets:

| Dataset           | Sentences      | Tokens            |
|-------------------|----------------|-------------------|
| DOCG v2           | 8.472.786      | 188.929.206       |
| El Periodico      | 6.483.106      | 145.591.906       |
| EuroParl          | 1.876.669      | 49.212.670        |
| WikiMatrix        | 1.421.077      | 34.902.039        |
| Wikimedia         | 335.955        | 8.682.025         |
| QED               | 71.867         | 1.079.705         |
| TED2020 v1        | 52.177         | 836.882           |
| CCMatrix v1       | 56.103.820     | 1.064.182.320     |
| MultiCCAligned v1 | 2.433.418      | 48.294.144        |
| ParaCrawl         | 15.327.808     | 334.199.408       |
| **Total**         | **92.578.683** | **1.875.910.305** |

### Training procedure

### Data preparation

 All datasets are concatenated and filtered using the [mBERT Gencata parallel filter](https://huggingface.co/projecte-aina/mbert-base-gencata) and cleaned using the clean-corpus-n.pl script from [moses](https://github.com/moses-smt/mosesdecoder), allowing sentences between 5 and 150 words.

 Before training, the punctuation is normalized using a modified version of the join-single-file.py script from [SoftCatalà](https://github.com/Softcatala/nmt-models/blob/master/data-processing-tools/join-single-file.py)


#### Tokenization

 All data is tokenized using sentencepiece, with 50 thousand token sentencepiece model  learned from the combination of all filtered training data. This model is included.  

#### Hyperparameters

The model is based on the Transformer-XLarge proposed by [Subramanian et al.](https://aclanthology.org/2021.wmt-1.18.pdf)
The following hyperparamenters were set on the Fairseq toolkit:

| Hyperparameter                     | Value                            |
|------------------------------------|----------------------------------|
| Architecture                       | transformer_vaswani_wmt_en_de_bi |
| Embedding size                     | 1024                             |
| Feedforward size                   | 4096                             |
| Number of heads                    | 16                               |
| Encoder layers                     | 24                               |
| Decoder layers                     | 6                                |
| Normalize before attention         | True                             |
| --share-decoder-input-output-embed | True                             |
| --share-all-embeddings             | True                             |
| Effective batch size               | 96.000                           |
| Optimizer                          | adam                             |
| Adam betas                         | (0.9, 0.980)                     |
| Clip norm                          | 0.0                              |
| Learning rate                      | 1e-3                             |
| Lr. schedurer                      | inverse sqrt                     |
| Warmup updates                     | 4000                             |
| Dropout                            | 0.1                              |
| Label smoothing                    | 0.1                              |

The model was trained using shards of 10 million sentences, for a total of 8.000 updates. Weights were saved every 1000 updates and reported results are the average of the last 6 checkpoints. 

## Evaluation

### Variable and metrics

We use the BLEU score for evaluation on test sets: [Flores-101](https://github.com/facebookresearch/flores), [TaCon](https://elrc-share.eu/repository/browse/tacon-spanish-constitution-mt-test-set/84a96138b98611ec9c1a00155d02670628f3e6857b0f422abd82abc3795ec8c2/), [United Nations](https://zenodo.org/record/3888414#.Y33-_tLMIW0), [Cybersecurity](https://elrc-share.eu/repository/browse/cyber-mt-test-set/2bd93faab98c11ec9c1a00155d026706b96a490ed3e140f0a29a80a08c46e91e/), [wmt19 biomedical test set](), [wmt13 news test set](https://elrc-share.eu/repository/browse/catalan-wmt2013-machine-translation-shared-task-test-set/84a96139b98611ec9c1a00155d0267061a0aa1b62e2248e89aab4952f3c230fc/)

### Evaluation results

Below are the evaluation results on the machine translation from Spanish to Catalan compared to [Softcatalà](https://www.softcatala.org/) and [Google Translate](https://translate.google.es/?hl=es):

| Test set             | SoftCatalà | Google Translate |mt-plantl-es-ca|
|----------------------|------------|------------------|---------------|
| Spanish Constitution | **63,6**   | 61,7             | 63,0          |
| United Nations       | 73,8       | 74,8             | **74,9**      |
| Flores 101 dev       | 22         | **23,1**         | 22,5          |
| Flores 101 devtest   | 22,7       | **23,6**         | 23,1          |
| Cybersecurity        | 61,4       | **69,5**         | 67,3          |
| wmt 19 biomedical    | 60,2       | 59,7             | **60,6**      |
| wmt 13 news          | 21,3       | **22,4**         | 22,0          |
| Average              | 46,4       | **47,8**         | 47,6          |

## Additional information

### Author
Text Mining Unit (TeMU) at the Barcelona Supercomputing Center ([email protected])

### Contact information
For further information, send an email to <[email protected]>

### Copyright
Copyright by the Spanish State Secretariat for Digitalization and Artificial Intelligence (SEDIA) (2022)

### Licensing information
This work is licensed under a [Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0)

### Funding
This work was funded by the Spanish State Secretariat for Digitalization and Artificial Intelligence (SEDIA)

### Disclaimer

<details>
<summary>Click to expand</summary>

The models published in this repository are intended for a generalist purpose and are available to third parties. These models may have bias and/or any other undesirable distortions.

When third parties, deploy or provide systems and/or services to other parties using any of these models (or using systems based on these models) or become users of the models, they should note that it is their responsibility to mitigate the risks arising from their use and, in any event, to comply with applicable regulations, including regulations regarding the use of Artificial Intelligence.

In no event shall the owner of the models (SEDIA – State Secretariat for Digitalization and Artificial Intelligence) nor the creator (BSC – Barcelona Supercomputing Center) be liable for any results arising from the use made by third parties of these models.


Los modelos publicados en este repositorio tienen una finalidad generalista y están a disposición de terceros. Estos modelos pueden tener sesgos y/u otro tipo de distorsiones indeseables.

Cuando terceros desplieguen o proporcionen sistemas y/o servicios a otras partes usando alguno de estos modelos (o utilizando sistemas basados en estos modelos) o se conviertan en usuarios de los modelos, deben tener en cuenta que es su responsabilidad mitigar los riesgos derivados de su uso y, en todo caso, cumplir con la normativa aplicable, incluyendo la normativa en materia de uso de inteligencia artificial.

En ningún caso el propietario de los modelos (SEDIA – Secretaría de Estado de Digitalización e Inteligencia Artificial) ni el creador (BSC – Barcelona Supercomputing Center) serán responsables de los resultados derivados del uso que hagan terceros de estos modelos.
</details>