Sentence-aware Contrastive Learning for Open-Domain Passage Retrieval

TL;DR

This paper proposes a sentence-aware contrastive learning approach for open-domain passage retrieval, focusing on finer granularity to better capture passage representations and reduce internal conflicts, leading to improved retrieval performance.

Contribution

It introduces an in-passage negative sampling strategy based on sentence-level granularity to enhance passage representation learning.

Findings

Effective on three benchmark datasets.

Improves performance especially with severe internal conflicts.

Demonstrates good transferability across datasets.

Abstract

Training dense passage representations via contrastive learning has been shown effective for Open-Domain Passage Retrieval (ODPR). Existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining. However, these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity. This work thus presents a refined model on the basis of a smaller granularity, contextual sentences, to alleviate the concerned conflicts. In detail, we introduce an in-passage negative sampling strategy to encourage a diverse generation of sentence representations within the same passage. Experiments on three benchmark datasets verify the efficacy of our method, especially on datasets where conflicts are severe. Extensive experiments further present good transferability of our method across datasets.