Explicit Alignment and Many-to-many Entailment Based Reasoning for Conversational Machine Reading

Yangyang Luo, Shiyu Tian, Caixia Yuan, Xiaojie Wang

Submission Type: Regular Long Paper

Submission Track: Dialogue and Interactive Systems

Keywords: Conversational Machine Reading, Task-oriented dialogue, Textual Entailment

Abstract: Conversational Machine Reading (CMR) requires answering a user's initial question through multi-turn dialogue interactions based on a given document. Although there exist many effective methods, they largely neglected the alignment between the $\textit{document}$ and the $\textit{user-provided information}$, which significantly affects the intermediate decision-making and subsequent follow-up question generation. To address this issue, we propose a pipeline framework that (1) aligns the aforementioned two sides in an explicit way, (2) makes decisions using a lightweight many-to-many entailment reasoning module, and (3) directly generates follow-up questions based on the document and previously asked questions. Our proposed method achieves state-of-the-art in micro-accuracy and ranks the first place on the public leaderboard of the CMR benchmark dataset ShARC.

Submission Number: 3588