A Bag of Tricks for Dialogue Summarization

TL;DR

This paper introduces techniques to improve dialogue summarization by addressing speaker differentiation, negation, reasoning, and informal language, using pretrained models and multi-task learning, resulting in better performance over existing baselines.

Contribution

It proposes novel methods like speaker substitution, negation highlighting, and in-domain pretraining to enhance dialogue summarization with pretrained sequence-to-sequence models.

Findings

Improved summarization accuracy over strong baselines

Effective handling of speaker and negation challenges

Benefits from multi-task learning and domain-specific pretraining

Abstract

Dialogue summarization comes with its own peculiar challenges as opposed to news or scientific articles summarization. In this work, we explore four different challenges of the task: handling and differentiating parts of the dialogue belonging to multiple speakers, negation understanding, reasoning about the situation, and informal language understanding. Using a pretrained sequence-to-sequence language model, we explore speaker name substitution, negation scope highlighting, multi-task learning with relevant tasks, and pretraining on in-domain data. Our experiments show that our proposed techniques indeed improve summarization performance, outperforming strong baselines.