m2m100 fine-tuned on Softcatalà's parallel Catalan-German dataset for machine translation
Table of Contents
<details> <summary>Click to expand</summary>
- Model description
- Intended uses and limitations
- How to Use
- Training
- Evaluation
- Additional information
</details>
Model description
This model was obtained by fine-tuning the m2m100_418M model on a De-Ca machine translation task with the Softcatalà Catalan-German parallel corpus dataset, with sentences deduplicated and filtered by the GEnCaTa quality filter. We also evaluate it on a general-domain multilingual testset Flores-200 and WMT13.
Intended uses and limitations
You can use this model for machine translation from German to Catalan.
How to use
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("projecte-aina/2m100-418M-ft-de-ca")
model = AutoModelForSeq2SeqLM.from_pretrained("projecte-aina/2m100-418M-ft-de-ca")
Training
Training data
As a data for fine-tuning we used the Softcatalà Catalan-German parallel corpus dataset, with sentences deduplicated and filtered by the GEnCaTa quality filter.
Training procedure
Tokenization
The original m2m100_418M model's sentencepiece tokenizer was used.
Hyperparameters
The model was trained for 2 epochs with the default parameters and \(LR = 2\mathrm{e}{-5}\).
Evaluation
Variable and metrics
We use the BLEU score for evaluation on test sets: Flores-200 and WMT13.
Evaluation results
Below are the evaluation results on the machine translation from German to Catalan compared with the original m2m100 on a testset: Flores-200.
| Test set | Model | BLEU | TER | METEOR | chrF |
|---|---|---|---|---|---|
| Flores-200 | m2m100 | 26.6 | 63.1 | 54.0 | 53.5 |
| m2m100-418M-ft-de-ca | 28.5 | 60.7 | 55.9 | 55.9 | |
| WMT13 | m2m100 | 21.8 | 72.8 | 48.0 | 53.5 |
| m2m100-418M-ft-de-ca | 22.9 | 71.0 | 49.5 | 54.1 |
Additional information
Licensing information
Funding
This work was funded by the [Departament de la Vicepresidència i de Polítiques Digitals i Territori de la Generalitat de Catalunya](https://politiquesdigitals.gencat.cat/ca/inici/index.html#googtrans(ca|en) within the framework of Projecte AINA.
Citation Information
@article{garriga2022catalan,
title={A Catalan-German machine translation system based on the M2M-100 multilingual model},
author={Garriga Riba, Pol},
year={2022},
url={https://repositori.upf.edu/bitstream/handle/10230/54301/GarrigaRiba_2022.pdf?sequence=1&isAllowed=y}
}
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 and creator of the models (BSC – Barcelona Supercomputing Center) be liable for any results arising from the use made by third parties of these models.
