projecte-aina / aina-translator-gl-ca

Projecte Aina’s Galician-Catalan machine translation model

Model description

This model was trained from scratch using the Fairseq toolkit on a combination of Galician-Catalan datasets totalling 10.017.995 sentence pairs. 4.267.995 sentence pairs were parallel data collected from the web while the remaining 5.750.000 sentence pairs were parallel synthetic data created using the GL-ES translator of Proxecto Nós . The model was evaluated on the Flores, TaCon and NTREX evaluation datasets.

Intended uses and limitations

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

How to use

Usage

Required libraries:

pip install ctranslate2 pyonmttok

Translate a sentence using python

import ctranslate2
import pyonmttok
from huggingface_hub import snapshot_download
model_dir = snapshot_download(repo_id="projecte-aina/aina-translator-gl-ca", revision="main")
tokenizer=pyonmttok.Tokenizer(mode="none", sp_model_path = model_dir + "/spm.model")
tokenized=tokenizer.tokenize("Benvido ao proxecto Ilenia.")
translator = ctranslate2.Translator(model_dir)
translated = translator.translate_batch([tokenized[0]])
print(tokenizer.detokenize(translated[0][0]['tokens']))

Limitations and bias

At the time of submission, no measures have been taken to estimate the bias and toxicity embedded in the model. However, we are well aware that our models may be biased. We intend to conduct research in these areas in the future, and if completed, this model card will be updated.

Training

Training data

The Catalan-Galician data is a combination of publicly available bilingual datasets collected from the web. These datasets were concatenated before filtering to avoid intra-dataset duplicates and the final size was 4.267.995. Additional 5.750.000 sentence pairs of synthetic parallel data were created from a random sampling of the Projecte Aina ES-CA corpus .

Training procedure

Data preparation

All datasets are deduplicated and filtered to remove any sentence pairs with a cosine similarity of less than 0.75. This is done using sentence embeddings calculated using LaBSE . The filtered datasets are then concatenated to form a final corpus of 10.017.995 and before training the punctuation is normalized using a modified version of the join-single-file.py script from SoftCatalà

Tokenization

All data is tokenized using sentencepiece, with a 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. The following hyperparameters were set on the Fairseq toolkit:

The model was trained for 24.000 updates on the parallel data collected from the web. This data was then concatenated with the synthetic parallel data and training continued for a total of 34.000 updates. Weights were saved every 1000 updates and reported results are the average of the last 4 checkpoints.

Evaluation

Variable and metrics

We use the BLEU score for evaluation on test sets: Flores-200 , TaCon and NTREX .

Evaluation results

Below are the evaluation results on the machine translation from Galician to Catalan compared to Google Translate , M2M100 1.2B , NLLB 200 3.3B and NLLB-200's distilled 1.3B variant :

Additional information

Author

The Language Technologies Unit from Barcelona Supercomputing Center.

Contact

For further information, please send an email to [email protected] .

Copyright

Copyright(c) 2023 by Language Technologies Unit, Barcelona Supercomputing Center.

License

Apache License, Version 2.0

Funding

This work is funded by the Ministerio para la Transformación Digital y de la Función Pública - Funded by EU – NextGenerationEU within the framework of the project ILENIA with reference 2022/TL22/00215337, 2022/TL22/00215336, 2022/TL22/00215335 y 2022/TL22/00215334

Disclaimer