Self- and Pseudo-self-supervised Prediction of Speaker and Key-utterance for Multi-party Dialogue Reading Comprehension

TL;DR

This paper introduces a novel self- and pseudo-self-supervised approach for multi-party dialogue reading comprehension that models speaker and key-utterance information without requiring manual labels, improving performance on benchmark datasets.

Contribution

It proposes labor-free self- and pseudo-self-supervised tasks to implicitly model speaker flows and salient dialogue clues, reducing reliance on complex modules and labeled data.

Findings

Outperforms competitive baselines and state-of-the-art models

Effectively captures speaker information without manual labels

Improves understanding of long multi-party dialogues

Abstract

Multi-party dialogue machine reading comprehension (MRC) brings tremendous challenge since it involves multiple speakers at one dialogue, resulting in intricate speaker information flows and noisy dialogue contexts. To alleviate such difficulties, previous models focus on how to incorporate these information using complex graph-based modules and additional manually labeled data, which is usually rare in real scenarios. In this paper, we design two labour-free self- and pseudo-self-supervised prediction tasks on speaker and key-utterance to implicitly model the speaker information flows, and capture salient clues in a long dialogue. Experimental results on two benchmark datasets have justified the effectiveness of our method over competitive baselines and current state-of-the-art models.