C-MORE: Pretraining to Answer Open-Domain Questions by Consulting Millions of References

TL;DR

This paper introduces C-MORE, a pretraining approach for open-domain QA systems that leverages millions of references from Wikipedia to create high-quality training triplets, significantly improving retriever and reader performance.

Contribution

The paper presents a novel method to automatically construct large-scale, high-quality QA triplets from Wikipedia references for pretraining retriever and reader components.

Findings

Pretrained retriever improves top-20 accuracy by 2-10%.

Pretrained reader enhances overall system performance by up to 4% in exact match.

The approach effectively covers diverse domains and aligns well with downstream tasks.

Abstract

We consider the problem of pretraining a two-stage open-domain question answering (QA) system (retriever + reader) with strong transfer capabilities. The key challenge is how to construct a large amount of high-quality question-answer-context triplets without task-specific annotations. Specifically, the triplets should align well with downstream tasks by: (i) covering a wide range of domains (for open-domain applications), (ii) linking a question to its semantically relevant context with supporting evidence (for training the retriever), and (iii) identifying the correct answer in the context (for training the reader). Previous pretraining approaches generally fall short of one or more of these requirements. In this work, we automatically construct a large-scale corpus that meets all three criteria by consulting millions of references cited within Wikipedia. The well-aligned pretraining signals benefit both the retriever and the reader significantly. Our pretrained retriever leads to 2%-10% absolute gains in top-20 accuracy. And with our pretrained reader, the entire system improves by up to 4% in exact match.