This website requires JavaScript.

STRUDEL: Structured Dialogue Summarization for Dialogue Comprehension

Borui WangChengcheng FengArjun Nair ...+5 Dragomir Radev
Dec 2022
摘要
Abstractive dialogue summarization has long been viewed as an importantstandalone task in natural language processing, but no previous work hasexplored the possibility of whether abstractive dialogue summarization can alsobe used as a means to boost an NLP system's performance on other importantdialogue comprehension tasks. In this paper, we propose a novel type ofdialogue summarization task - STRUctured DiaLoguE Summarization - that can helppre-trained language models to better understand dialogues and improve theirperformance on important dialogue comprehension tasks. We further collect humanannotations of STRUDEL summaries over 400 dialogues and introduce a new STRUDELdialogue comprehension modeling framework that integrates STRUDEL into agraph-neural-network-based dialogue reasoning module over transformer encoderlanguage models to improve their dialogue comprehension abilities. In ourempirical experiments on two important downstream dialogue comprehension tasks- dialogue question answering and dialogue response prediction - we show thatour STRUDEL dialogue comprehension model can significantly improve the dialoguecomprehension performance of transformer encoder language models.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答