Linear Cross-Lingual Mapping of Sentence Embeddings

TL;DR

This paper proposes a simple linear approach to improve multilingual sentence embeddings by ensuring translation invariance and analyzing orthogonality deviations to better preserve sentence semantics across languages.

Contribution

It introduces a linear cross-lingual mapping method and evaluates orthogonality deviations to enhance multilingual sentence embedding quality.

Findings

Linear mapping improves cross-lingual sentence similarity

Orthogonality deviations indicate embedding deficiencies

Translation invariance correlates with semantic preservation

Abstract

Semantics of a sentence is defined with much less ambiguity than semantics of a single word, and we assume that it should be better preserved by translation to another language. If multilingual sentence embeddings intend to represent sentence semantics, then the similarity between embeddings of any two sentences must be invariant with respect to translation. Based on this suggestion, we consider a simple linear cross-lingual mapping as a possible improvement of the multilingual embeddings. We also consider deviation from orthogonality conditions as a measure of deficiency of the embeddings.