Improving Zero-Shot Multi-Lingual Entity Linking

  • 2021-04-16 12:50:07
  • Elliot Schumacher, James Mayfield, Mark Dredze
  • 2

Abstract

Entity linking -- the task of identifying references in free text to relevantknowledge base representations -- often focuses on single languages. Weconsider multilingual entity linking, where a single model is trained to linkreferences to same-language knowledge bases in several languages. We propose aneural ranker architecture, which leverages multilingual transformerrepresentations of text to be easily applied to a multilingual setting. We thenexplore how a neural ranker trained in one language (e.g. English) transfers toan unseen language (e.g. Chinese), and find that while there is a consistentbut not large drop in performance. How can this drop in performance bealleviated? We explore adding an adversarial objective to force our model tolearn language-invariant representations. We find that using this approachimproves recall in several datasets, often matching the in-languageperformance, thus alleviating some of the performance loss occurring fromzero-shot transfer.

 

Quick Read (beta)

loading the full paper ...