Action based Network for Conversation Question Reformulation

TL;DR

This paper introduces an action-based method for conversation question reformulation that improves understanding of co-references and ellipsis, enhancing question completeness and accuracy in dialogue systems.

Contribution

It proposes a novel action-based framework for question reformulation that explicitly locates and resolves co-references and ellipsis in conversational questions.

Findings

Achieved 3.9% improvement in EM on Restoration-200K dataset.

Improved ROUGE-L score by 1.0% on the same dataset.

Effective in both English and Chinese utterance rewrite tasks.

Abstract

Conversation question answering requires the ability to interpret a question correctly. Current models, however, are still unsatisfactory due to the difficulty of understanding the co-references and ellipsis in daily conversation. Even though generative approaches achieved remarkable progress, they are still trapped by semantic incompleteness. This paper presents an action-based approach to recover the complete expression of the question. Specifically, we first locate the positions of co-reference or ellipsis in the question while assigning the corresponding action to each candidate span. We then look for matching phrases related to the candidate clues in the conversation context. Finally, according to the predicted action, we decide whether to replace the co-reference or supplement the ellipsis with the matched information. We demonstrate the effectiveness of our method on both English and Chinese utterance rewrite tasks, improving the state-of-the-art EM (exact match) by 3.9\% and ROUGE-L by 1.0\% respectively on the Restoration-200K dataset.