A Relaxed Matching Procedure for Unsupervised BLI

TL;DR

This paper introduces a relaxed matching procedure for unsupervised Bilingual Lexicon Induction that improves translation pair accuracy by reducing constraints and leveraging bidirectional embedding alignment, outperforming previous methods.

Contribution

It proposes a novel relaxed matching approach and demonstrates the benefits of bidirectional embedding alignment in unsupervised BLI.

Findings

Significant improvement over previous unsupervised methods

Relaxed matching reduces counterintuitive pairings

Bidirectional alignment enhances translation accuracy

Abstract

Recently unsupervised Bilingual Lexicon Induction (BLI) without any parallel corpus has attracted much research interest. One of the crucial parts in methods for the BLI task is the matching procedure. Previous works impose a too strong constraint on the matching and lead to many counterintuitive translation pairings. Thus, We propose a relaxed matching procedure to find a more precise matching between two languages. We also find that aligning source and target language embedding space bidirectionally will bring significant improvement. We follow the previous iterative framework to conduct experiments. Results on standard benchmark demonstrate the effectiveness of our proposed method, which substantially outperforms previous unsupervised methods.