Mixed-modality Representation Learning and Pre-training for Joint Table-and-Text Retrieval in OpenQA

TL;DR

This paper introduces OTTeR, a novel joint table-and-text retriever for open-domain QA that leverages mixed-modality representation learning and synthetic pre-training to improve evidence retrieval accuracy.

Contribution

The paper proposes a new retriever model with enhanced mixed-modality learning and pre-training strategies, addressing data sparsity and modality discrepancy issues.

Findings

OTTeR significantly outperforms previous methods on OTT-QA dataset.

The proposed mechanisms effectively improve retrieval performance.

The system achieves state-of-the-art results in open-domain QA.

Abstract

Retrieving evidences from tabular and textual resources is essential for open-domain question answering (OpenQA), which provides more comprehensive information. However, training an effective dense table-text retriever is difficult due to the challenges of table-text discrepancy and data sparsity problem. To address the above challenges, we introduce an optimized OpenQA Table-Text Retriever (OTTeR) to jointly retrieve tabular and textual evidences. Firstly, we propose to enhance mixed-modality representation learning via two mechanisms: modality-enhanced representation and mixed-modality negative sampling strategy. Secondly, to alleviate data sparsity problem and enhance the general retrieval ability, we conduct retrieval-centric mixed-modality synthetic pre-training. Experimental results demonstrate that OTTeR substantially improves the performance of table-and-text retrieval on the OTT-QA dataset. Comprehensive analyses examine the effectiveness of all the proposed mechanisms. Besides, equipped with OTTeR, our OpenQA system achieves the state-of-the-art result on the downstream QA task, with 10.1% absolute improvement in terms of the exact match over the previous best system. All the code and data are available at https://github.com/Jun-jie-Huang/OTTeR.