Unsupervised Neural Single-Document Summarization of Reviews via Learning Latent Discourse Structure and its Ranking

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 2142–2152 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 2142 Unsupervised Neural Single-Document Summarization of Reviews via Learning Latent Discourse Structure and its Ranking Masaru Isonuma1 Junichiro Mori1,2 Ichiro Sakata1 1The University of Tokyo 2RIKEN {isonuma, isakata}@ipr-ctr.t.u-tokyo.ac.jp mori@mi.u-tokyo.ac.jp Abstract This paper focuses on the end-to-end abstrac- tive summarization of a single product review without supervision. We assume that a review can be described as a discourse tree, in which the summary is the root, and the child sen- tences explain their parent in detail. By re- cursively estimating a parent from its children, our model learns the latent discourse tree with- out an external parser and generates a concise summary. We also introduce an architecture that ranks the importance of each sentence on the tree to support summary generation focus- ing on the main review point. The experimen- tal results demonstrate that our model is com- petitive with or outperforms other unsuper- vised approaches. In particular, for relatively long reviews, it achieves a competitive or bet- ter performance than supervised models. The induced tree shows that the child sentences provide additional information about their par- ent, and the generated summary abstracts the entire review. 1 Introduction The need for automatic document summarization is widely increasing because of the vast amounts of online textual data that continue to grow. As for product reviews on E-commerce websites, succinct summaries allow both customers and manufacturers to obtain large numbers of opin- ions (Liu and Zhang, 2012). Under these cir- cumstances, supervised neural network models have achieved wide success, using a large number of reference summaries (Wang and Ling, 2016; Ma et al., 2018). However, a model trained on these summaries cannot be adopted in other do- mains, as salient phrases are not common across domains. It requires a significant cost to pre- pare large volumes of references for each domain (Isonuma et al., 2017). An unsupervised approach is a possible solu- tion to such a problem. Previously, unsupervised learning has been widely applied to extractive ap- proaches (Radev et al., 2004; Mihalcea and Tarau, 2004). As mentioned in (Carenini et al., 2013; Gerani et al., 2014), extractive approaches often fail to provide an overview of the reviews, while abstractive ones successfully condense an entire review via paraphrasing and generalization. Our work focuses on the one-sentence abstractive sum- marization of a single-review without supervision. The difficulties of unsupervised abstractive summarization are two-fold: obtaining the repre- sentation of the summaries, and learning a lan- guage model to decode them. As an unsuper- vised approach for multiple reviews, Chu and Liu (2018) regarded the mean of the document embed- dings as the summary, while learning a language model via the reconstruction of each review. By contrast, such an approach cannot be extended to a single-review directly, because it also condenses including trivial or redundant sentences (its perfor- mance is demonstrated in Section 4.4). To overcome these problems, we …

Content truncated to 500 words