Spanish legal domain Language Model ⚖️

This repository contains the page for two main resources for the Spanish legal domain:

• A RoBERTa model: https://huggingface.co/PlanTL-GOB-ES/RoBERTalex
• FastText embeddings: https://zenodo.org/record/5036147
• Legal corpora: https://zenodo.org/record/5495529

The repository and the pre-print will be updated with larger models, evaluations, etcetera.

Why ❓

There are few models trained for the Spanish language. Some of the models have been trained with a low resource, unclean corpora. The ones derived from the Spanish National Plan for Language Technologies are proficient solving several tasks and have been trained using large scale clean corpora. However, the Spanish Legal domain language could be think of an independent language on its own. We therefore created a Spanish Legal model from scratch trained exclusively on legal corpora.

Evaluation ✅

Work in progress.

Corpora 📃

Usage example ⚗️

You can train your model for different downstream tasks using the scripts that Hugging Face provides (Name Entity Recognition, GLUE tasks and others)

from transformers import AutoModelForMaskedLM
from transformers import AutoTokenizer, FillMaskPipeline
from pprint import pprint
tokenizer_hf = AutoTokenizer.from_pretrained('PlanTL-GOB-ES/RoBERTalex')
model = AutoModelForMaskedLM.from_pretrained('PlanTL-GOB-ES/RoBERTalex')
model.eval()
pipeline = FillMaskPipeline(model, tokenizer_hf)
text = f"¡Hola <mask>!"
res_hf = pipeline(text)
pprint([r['token_str'] for r in res_hf])

Cite 📣

If this work is helpful, please cite it:

@misc{gutierrezfandino2021legal,
      title={Spanish Legalese Language Model and Corpora}, 
      author={Asier Gutiérrez-Fandiño and Jordi Armengol-Estapé and Aitor Gonzalez-Agirre and Marta Villegas},
      year={2021},
      eprint={2110.12201},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Contact 📧

📋 We are interested in (1) extending our corpora to make larger models (2) evaluate/train the model in other tasks.

For questions regarding this work, contact Asier Gutiérrez-Fandiño ([email protected])