A Monolingual Approach to Contextualized Word Embeddings for Mid-Resource Languages

TL;DR

This paper demonstrates that monolingual contextualized word embeddings trained on large, noisy web-crawled corpora can outperform Wikipedia-based embeddings and even multilingual models in tagging and parsing for mid-resource languages.

Contribution

It introduces monolingual ELMo embeddings trained on the OSCAR corpus for five mid-resource languages, showing improved performance over Wikipedia-based and multilingual models.

Findings

OSCAR-based embeddings outperform Wikipedia-based embeddings.

Embeddings trained on OSCAR match or exceed state-of-the-art in tagging and parsing.

Larger, diverse corpora benefit monolingual embeddings more than multilingual architectures.

Abstract

We use the multilingual OSCAR corpus, extracted from Common Crawl via language classification, filtering and cleaning, to train monolingual contextualized word embeddings (ELMo) for five mid-resource languages. We then compare the performance of OSCAR-based and Wikipedia-based ELMo embeddings for these languages on the part-of-speech tagging and parsing tasks. We show that, despite the noise in the Common-Crawl-based OSCAR data, embeddings trained on OSCAR perform much better than monolingual embeddings trained on Wikipedia. They actually equal or improve the current state of the art in tagging and parsing for all five languages. In particular, they also improve over multilingual Wikipedia-based contextual embeddings (multilingual BERT), which almost always constitutes the previous state of the art, thereby showing that the benefit of a larger, more diverse corpus surpasses the cross-lingual benefit of multilingual embedding architectures.