Analogy Training Multilingual Encoders

Authors

  • Nicolas Garneau Université Laval
  • Mareike Hartmann University of Copenhagen
  • Anders Sandholm Google Research
  • Sebastian Ruder DeepMind
  • Ivan Vulić University of Cambridge
  • Anders Søgaard University of Copenhagen

Keywords:

Language Models

Abstract

Language encoders encode words and phrases in ways that capture their local semantic relatedness, but are known to be globally inconsistent. Global inconsistency can seemingly be corrected for, in part, by leveraging signals from knowledge bases, but previous results are partial and limited to monolingual English encoders. We extract a large-scale multilingual, multi-word analogy dataset from Wikidata for diagnosing and correcting for global inconsistencies, and then implement a four-way Siamese BERT architecture for grounding multilingual BERT (mBERT) in Wikidata through analogy training. We show that analogy training not only improves the global consistency of mBERT, as well as the isomorphism of language-specific subspaces, but also leads to consistent gains on downstream tasks such as bilingual dictionary induction and sentence retrieval.

Downloads

Published

2021-05-18

How to Cite

Garneau, N., Hartmann, M., Sandholm, A., Ruder, S., Vulić, I., & Søgaard, A. (2021). Analogy Training Multilingual Encoders. Proceedings of the AAAI Conference on Artificial Intelligence, 35(14), 12884-12892. Retrieved from https://ojs.aaai.org/index.php/AAAI/article/view/17524

Issue

Section

AAAI Technical Track on Speech and Natural Language Processing I