jramompichel's picture
Update README.md
7a6fd74
|
raw
history blame
No virus
5.52 kB
metadata
license: mit
language:
  - gl
metrics:
  - bleu (Gold1): 79.6
  - bleu (Gold2): 43.3
  - bleu (Flores): 21.8
  - bleu (Test-suite): 74.3

Descrici贸n do Modelo / Model description

Modelo feito con OpenNMT para o par espa帽ol-galego utilizando unha arquitectura transformer.

Model developed with OpenNMT for the Spanish-Galician pair using a transformer architecture.

Como traducir / How to translate

  • 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 traduci贸n estar谩 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 aut茅nticos e sint茅ticos. Os primeiros son corpora de traduci贸ns feitas directamente por tradutores humanos. Os segundos son corpora de traduci贸ns espa帽ol-portugu茅s e ingl茅s-portugu茅s, que convertemos en espa帽ol-galego e ingl茅s-galego a trav茅s da traduci贸n autom谩tica portugu茅s-galego con Opentrad/Apertium e transliteraci贸n para palabras f贸ra de vocabul谩rio.

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 trav茅s 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

Hiperpar谩metros / Hyper-parameters

Os par谩metros 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

Avaliaci贸n / Evaluation

A avaliaci贸n BLEU dos modelos 茅 feita cunha mistura de tests desenvolvidos internamente (gold1, gold2, test-suite) con outros datasets dispon铆beis en galego (Flores).

The BLEU 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 License

Copyright (c) 2023 Proxecto N贸s

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Financiamento / Funding

Esta investigaci贸n foi financiada polo proxecto "N贸s: o galego na sociedade e econom铆a da intelixencia artificial", resultado dun acordo entre a Xunta de Galicia e a Universidade de Santiago de Compostela, o que resultou no subsidio ED431G2019/04 da Consellar铆a de Educaci贸n, Universidade e Formaci贸n Profesional da Galiza, e polo Fondo Europeo de Desenvolvemento Rexional (programa ERDF/FEDER), e Grupos de Referencia: ED431C 2020/21.

This research was funded by the project "N贸s: 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