Multilingual Information Retrieval with a Monolingual Knowledge Base

TL;DR

This paper introduces a novel fine-tuning approach for multilingual embedding models using weighted sampling for contrastive learning, enabling effective cross-language information retrieval with a monolingual knowledge base.

Contribution

It proposes a new weighted sampling strategy for contrastive learning that improves multilingual retrieval performance and is applicable across various languages and code-switching scenarios.

Findings

Performance gains of up to 31.03% in MRR

Performance gains of up to 33.98% in Recall@3

Method is language-agnostic and versatile

Abstract

Multilingual information retrieval has emerged as powerful tools for expanding knowledge sharing across languages. On the other hand, resources on high quality knowledge base are often scarce and in limited languages, therefore an effective embedding model to transform sentences from different languages into a feature vector space same as the knowledge base language becomes the key ingredient for cross language knowledge sharing, especially to transfer knowledge available in high-resource languages to low-resource ones. In this paper we propose a novel strategy to fine-tune multilingual embedding models with weighted sampling for contrastive learning, enabling multilingual information retrieval with a monolingual knowledge base. We demonstrate that the weighted sampling strategy produces performance gains compared to standard ones by up to 31.03\% in MRR and up to 33.98\% in Recall@3. Additionally, our proposed methodology is language agnostic and applicable for both multilingual and code switching use cases.