Improving Multi-lingual Alignment Through Soft Contrastive Learning

TL;DR

This paper introduces a novel soft contrastive learning method for multi-lingual sentence embeddings, leveraging a pre-trained mono-lingual model to improve alignment and outperform existing models in cross-lingual tasks.

Contribution

It proposes a new soft contrastive loss approach for multi-lingual embedding alignment using sentence similarity from mono-lingual models, enhancing performance over traditional methods.

Findings

Outperforms conventional contrastive loss with hard labels.

Achieves superior results on bitext mining and STS benchmarks.

Outperforms LaBSE on Tatoeba dataset.

Abstract

Making decent multi-lingual sentence representations is critical to achieve high performances in cross-lingual downstream tasks. In this work, we propose a novel method to align multi-lingual embeddings based on the similarity of sentences measured by a pre-trained mono-lingual embedding model. Given translation sentence pairs, we train a multi-lingual model in a way that the similarity between cross-lingual embeddings follows the similarity of sentences measured at the mono-lingual teacher model. Our method can be considered as contrastive learning with soft labels defined as the similarity between sentences. Our experimental results on five languages show that our contrastive loss with soft labels far outperforms conventional contrastive loss with hard labels in various benchmarks for bitext mining tasks and STS tasks. In addition, our method outperforms existing multi-lingual embeddings including LaBSE, for Tatoeba dataset. The code is available at https://github.com/YAI12xLinq-B/IMASCL