Can Question Rewriting Help Conversational Question Answering?

TL;DR

This paper investigates whether question rewriting can improve conversational question answering by using reinforcement learning, but finds that it performs similarly to baseline methods and faces significant challenges.

Contribution

It introduces a reinforcement learning approach integrating question rewriting with CQA without needing dedicated QR datasets, and analyzes its limitations.

Findings

RL approach performs on par with baseline

QR's potential in CQA remains challenging

Difficulty in effectively exploiting QR for CQA

Abstract

Question rewriting (QR) is a subtask of conversational question answering (CQA) aiming to ease the challenges of understanding dependencies among dialogue history by reformulating questions in a self-contained form. Despite seeming plausible, little evidence is available to justify QR as a mitigation method for CQA. To verify the effectiveness of QR in CQA, we investigate a reinforcement learning approach that integrates QR and CQA tasks and does not require corresponding QR datasets for targeted CQA. We find, however, that the RL method is on par with the end-to-end baseline. We provide an analysis of the failure and describe the difficulty of exploiting QR for CQA.