Gromov-Wasserstein Alignment of Word Embedding Spaces

TL;DR

This paper introduces a novel approach to align monolingual word embeddings across languages using Gromov-Wasserstein optimal transport, achieving competitive results without complex heuristics.

Contribution

It formulates the alignment as a Gromov-Wasserstein optimal transport problem, simplifying the process and reducing the need for tuning compared to existing methods.

Findings

Achieves performance comparable to state-of-the-art methods

Requires minimal tuning and heuristic post-processing

Efficient estimation of the OT objective

Abstract

Cross-lingual or cross-domain correspondences play key roles in tasks ranging from machine translation to transfer learning. Recently, purely unsupervised methods operating on monolingual embeddings have become effective alignment tools. Current state-of-the-art methods, however, involve multiple steps, including heuristic post-hoc refinement strategies. In this paper, we cast the correspondence problem directly as an optimal transport (OT) problem, building on the idea that word embeddings arise from metric recovery algorithms. Indeed, we exploit the Gromov-Wasserstein distance that measures how similarities between pairs of words relate across languages. We show that our OT objective can be estimated efficiently, requires little or no tuning, and results in performance comparable with the state-of-the-art in various unsupervised word translation tasks.