Leveraging Multi-lingual Positive Instances in Contrastive Learning to Improve Sentence Embedding

TL;DR

This paper introduces MPCL, a contrastive learning method that leverages multiple positive instances across languages to enhance multi-lingual sentence embeddings, resulting in improved cross-lingual transfer and downstream task performance.

Contribution

The paper proposes a novel contrastive learning approach, MPCL, that utilizes multiple positives for better multi-lingual sentence embeddings, especially in unseen languages.

Findings

MPCL improves retrieval, semantic similarity, and classification tasks.

Models trained with MPCL perform better in cross-lingual transfer.

Using multiple positives benefits learning across diverse languages.

Abstract

Learning multi-lingual sentence embeddings is a fundamental task in natural language processing. Recent trends in learning both mono-lingual and multi-lingual sentence embeddings are mainly based on contrastive learning (CL) among an anchor, one positive, and multiple negative instances. In this work, we argue that leveraging multiple positives should be considered for multi-lingual sentence embeddings because (1) positives in a diverse set of languages can benefit cross-lingual learning, and (2) transitive similarity across multiple positives can provide reliable structural information for learning. In order to investigate the impact of multiple positives in CL, we propose a novel approach, named MPCL, to effectively utilize multiple positive instances to improve the learning of multi-lingual sentence embeddings. Experimental results on various backbone models and downstream tasks demonstrate that MPCL leads to better retrieval, semantic similarity, and classification performances compared to conventional CL. We also observe that in unseen languages, sentence embedding models trained on multiple positives show better cross-lingual transfer performance than models trained on a single positive instance.