Composing Elementary Discourse Units in Abstractive Summarization

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6191–6196 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 6191 Composing Elementary Discourse Units in Abstractive Summarization Zhenwen Li and Wenhao Wu and Sujian Li Key Laboratory of Computational Linguistics(MOE) Department of Computer Science, Peking University {lizhenwen, waynewu, lisujian}@pku.edu.cn Abstract In this paper, we argue that elementary dis- course unit (EDU) is a more appropriate tex- tual unit of content selection than the sentence unit in abstractive summarization. To well han- dle the problem of composing EDUs into an informative and fluent summary, we propose a novel summarization method that first designs an EDU selection model to extract and group informative EDUs and then an EDU fusion model to fuse the EDUs in each group into one sentence. We also design the reinforcement learning mechanism to use EDU fusion results to reward the EDU selection action, boosting the final summarization performance. Experi- ments on CNN/Daily Mail have demonstrated the effectiveness of our model. 1 Introduction Abstractive summarization focuses on generating fluent and concise text from the original input doc- ument and has achieved considerable performance improvement with the rapid development of deep learning technology (See et al., 2017; Paulus et al., 2017; Celikyilmaz et al., 2018; Gehrmann et al., 2018). In abstractive summarization, the recently popular and practical paradigm usually generates summary sentences by independently compress- ing or rewriting each pre-extracted sentence, which is from the source documents (Chen and Bansal, 2018; Lebanoff et al., 2019). However, a single document sentence usually cannot provide enough information that a summary sentence expresses, which is supported by the re- cent study of Lebanoff et al. (2019). They show that a high percentage of summary sentences include in- formation from more than one document sentences, and composing a summary through only compress- ing sentences can cause performance degradation. Simultaneously, in contrast to the brevity require- ments of a summary, each document sentence usu- ally offers trivial details and expresses a relatively independent meaning, posing difficulty of combin- ing multiple sentences into one summary sentence. So we hope to seek a new summary composition unit which is more information-intensive and ele- mentary than sentence. In this paper, we choose to use Elementary Discourse Unit (EDU) as the summarization unit, which is first proposed from Rhetorical Structure Theory (Mann and Thompson, 1988) and defined as a clause. The finer granularity makes EDU more suitable than sentence to be the basic summary composition unit (Li et al., 2016). At the same time, benefited from the development of EDU seg- mentation technology, which can achieve a high accuracy of 94% (Wang et al., 2018), it is feasible to automatically obtain EDUs from the text. Next, the problems are: (1) which EDUs should be se- lected to compose a good summary? Moreover, (2) how to well assemble the selected EDUs into a fluent summary? To solve the problems above, we need to extract the information-intensive EDUs from the source documents and effectively fuse …

Content truncated to 500 words