license: mit
Descrición do Modelo / Model description
Modelo feito con OpenNMT para o par español-galego utilizando unha arquitectura transformer.
Model developed with OpenNMT for the Spanish-Galician pair using a transformer architecture.
Como utilizar
- Abrir terminal bash / Open bash terminal
- Instalar / Installing Python 3.9
- Instalar / Installing Open NMT toolkit v.2.2
- Traducir un input_text utilizando o modelo NOS-MT-es-gl co seguinte comando / Translating an input_text using the NOS-MT-en-gl model with the following command:
onmt_translate -src input_text -model NOS-MT-es-gl -output ./output_file.txt -replace_unk -phrase_table phrase_table-es-gl.txt -gpu 0
- O resultado da tradución estará no PATH indicado no flag -output / The result of the translation will be in the PATH indicated by the -output flag.
Adestramento / Training
No adestramento, utilizamos corpora auténticos e sintéticos. Os primeiros son corpora de traducións feitas directamente por tradutores humanos. Os segundos son corpora de traducións español-portugués e inglés-portugués, que convertemos en español-galego e inglés-galego a través da tradución automática portugués-galego con Opentrad/Apertium e transliteración para palabras fóra de vocabulário.
In the training we have used authentic and synthetic corpora. The former are corpora of translations directly produced by human translators. The latter are corpora of Spanish-Portuguese and English-Portuguese translations, which we have converted into Spanish-Galician and English-Galician by means of Portuguese-Galician translation with Opentrad/Apertium and transliteration for out-of-vocabulary words.
Procedemento de adestramento / Training process
Tokenization dos datasets feita co tokenizador de linguakit / Tokenization of the datasets made with linguakit tokeniser https://github.com/citiususc/Linguakit
O vocabulario para os modelos foi xerado a través do script / Vocabulary for the models was created by the script learn_bpe.py da open NMT
Usando o .yaml neste repositorio pode replicar o proceso de adestramento do seguinte xeito / Using the .yaml in this repository you can replicate the training process as follows
onmt_build_vocab -config bpe-es-gl_emb.yaml -n_sample 100000
onmt_train -config bpe-es-gl_emb.yaml
Hiperparámetros / Hyper-parameters
Os parámetros usados para o desenvolvimento do modelo poden ser consultados directamente no mesmo ficheiro .yaml bpe-es-gl_emb.yaml
The parameters used for the development of the model can be directly viewed in the same .yaml file bpe-es-gl_emb.yaml
Avaliación / Evaluation
A avalación dos modelos é feita cunha mistura de tests desenvolvidos internamente (gold1, gold2, test-suite) con outros datasets disponíbeis en galego (Flores).
The evaluation of the models is done by mixing internally developed tests (gold1, gold2, test-suite) with other datasets available in Galician (Flores).
| GOLD 1 | GOLD 2 | FLORES | TEST-SUITE |
|---|---|---|---|
| 79.6 | 43.3 | 21.8 | 74.3 |
Licenzas do Modelo / Licensing information
MIT
Financiamento / Funding
This research was funded by the project "Nós: Galician in the society and economy of artificial intelligence", agreement between Xunta de Galicia and University of Santiago de Compostela, and grant ED431G2019/04 by the Galician Ministry of Education, University and Professional Training, and the European Regional Development Fund (ERDF/FEDER program), and Groups of Reference: ED431C 2020/21.
Citation Information
@article{garriga2022catalan, title={}, author={}, year={2023}, url={} }