MarIA: Spanish Language Models

TL;DR

MarIA introduces a family of large, proficient Spanish language models including RoBERTa and GPT2 variants, trained on extensive Spanish web data, and demonstrates their superior performance on multiple NLP benchmarks.

Contribution

This work presents the largest and most proficient Spanish language models to date, along with new evaluation datasets, advancing NLP resources for Spanish.

Findings

MarIA models outperform existing Spanish models on various NLP tasks.

Models trained on 570GB of Spanish web data show high proficiency.

New extractive QA dataset created for evaluation.

Abstract

This work presents MarIA, a family of Spanish language models and associated resources made available to the industry and the research community. Currently, MarIA includes RoBERTa-base, RoBERTa-large, GPT2 and GPT2-large Spanish language models, which can arguably be presented as the largest and most proficient language models in Spanish. The models were pretrained using a massive corpus of 570GB of clean and deduplicated texts with 135 billion words extracted from the Spanish Web Archive crawled by the National Library of Spain between 2009 and 2019. We assessed the performance of the models with nine existing evaluation datasets and with a novel extractive Question Answering dataset created ex novo. Overall, MarIA models outperform the existing Spanish models across a variety of NLU tasks and training settings.