Unsupervised Extractive Opinion Summarization Using Sparse Coding

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 1209 - 1225 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Unsupervised Extractive Opinion Summarization Using Sparse Coding Somnath Basu Roy Chowdhury Chao Zhao Snigdha Chaturvedi {somnath, zhaochao, snigdha}@cs.unc.edu UNC Chapel Hill Abstract Opinion summarization is the task of auto- matically generating summaries that encapsu- late information from multiple user reviews. We present Semantic Autoencoder (SemAE) to perform extractive opinion summarization in an unsupervised manner. SemAE uses dic- tionary learning to implicitly capture semantic information from the review and learns a la- tent representation of each sentence over se- mantic units. A semantic unit is supposed to capture an abstract semantic concept. Our extractive summarization algorithm leverages the representations to identify representative opinions among hundreds of reviews. Se- mAE is also able to perform controllable sum- marization to generate aspect-specific sum- maries. We report strong performance on SPACE and AMAZON datasets, and perform experiments to investigate the functioning of our model. Our code is publicly available at https://github.com/brcsomnath/SemAE. 1 Introduction Opinion summarization is the task of automatically generating digests for an entity (e.g. a product, a hotel, a service, etc.), from user opinions in on- line forums. Automatic opinion summaries enable faster comparison, search, and better consumer feedback understanding (Hu and Liu, 2004; Pang, 2008; Medhat et al., 2014). Although there has been significant progress towards summarization (Rush et al., 2015; Nallapati et al., 2016; Cheng and Lapata, 2016; See et al., 2017; Narayan et al., 2018; Liu et al., 2018), existing approaches rely on human-annotated reference summaries, which are scarce for opinion summarization. For opinion summarization, human annotators need to read hun- dreds of reviews per entity across different sources for writing a summary, which may not be feasible. This lack of labeled training data has prompted a series of works to leverage unsupervised or weakly- supervised techniques for opinion summarization (Mei et al., 2007; Titov and McDonald, 2008; An- gelidis and Lapata, 2018a; Angelidis et al., 2021). Recent works in this direction have focused on per- forming opinion summarization in an abstractive setting (Coavoux et al., 2019; Isonuma et al., 2019; Bražinskas et al., 2020; Amplayo et al., 2021b; Iso et al., 2021; Wang and Wan, 2021). Abstrac- tive models are able to produce fluent summaries using novel phrases. However, they suffer from problems common in text generation like halluci- nation (Rohrbach et al., 2018), text degeneration (Holtzman et al., 2020), and topic drift (Sun et al., 2020). Also, these approaches have been evaluated on small scales (10 reviews per entity or fewer), which does not reveal their utility in the real world where there are hundreds of reviews per entity. To overcome these issues, another thread of works focuses on extractive opinion summarization, which creates summaries by selecting review sen- tences to reflect the popular opinions corresponding to an entity. A recently proposed extractive summa- rization approach is Quantized Transformer (QT) (Angelidis et al., 2021), which leverages vector quantization (van …

Content truncated to 500 words