Smoothing Dialogue States for Open Conversational Machine Reading

TL;DR

This paper introduces a gating strategy that smooths dialogue states within a single decoder to improve information transmission in open conversational machine reading, leading to state-of-the-art results.

Contribution

It proposes a novel gating approach that bridges decision making and question generation in a unified decoder for better performance.

Findings

Achieves new state-of-the-art results on OR-ShARC dataset.

Demonstrates improved dialogue state representation and information flow.

Outperforms existing pipeline and independent models.

Abstract

Conversational machine reading (CMR) requires machines to communicate with humans through multi-turn interactions between two salient dialogue states of decision making and question generation processes. In open CMR settings, as the more realistic scenario, the retrieved background knowledge would be noisy, which results in severe challenges in the information transmission. Existing studies commonly train independent or pipeline systems for the two subtasks. However, those methods are trivial by using hard-label decisions to activate question generation, which eventually hinders the model performance. In this work, we propose an effective gating strategy by smoothing the two dialogue states in only one decoder and bridge decision making and question generation to provide a richer dialogue state reference. Experiments on the OR-ShARC dataset show the effectiveness of our method, which achieves new state-of-the-art results.