Self-Supervised and Controlled Multi-Document Opinion Summarization

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 1646–1662 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 1646 Self-Supervised and Controlled Multi-Document Opinion Summarization Hady Elsahar,1 Maximin Coavoux,2 Matthias Gall´e,1 Jos Rozen1 {hady.elsahar,matthias.galle,jos.rozen}@naverlabs.com maximin.coavoux@univ-grenoble-alpes.fr 1Naver Labs Europe 2Univ. Grenoble Alpes, CNRS, Grenoble INP, LIG Abstract We address the problem of unsupervised ab- stractive summarization of collections of user generated reviews through self-supervision and control. We propose a self-supervised setup that considers an individual document as a target summary for a set of similar doc- uments. This setting makes training simpler than previous approaches by relying only on standard log-likelihood loss and mainstream models. We address the problem of halluci- nations through the use of control codes, to steer the generation towards more coherent and relevant summaries. Our benchmarks on two English datasets against graph-based and recent neural abstractive unsupervised models show that our proposed method generates sum- maries with a superior quality and relevance, as well as a high sentiment and topic align- ment with the input reviews. This is confirmed in our human evaluation which focuses explic- itly on the faithfulness of generated summaries. We also provide an ablation study showing the importance of the control setup in controlling hallucinations. 1 Introduction Recent progress in unsupervised methods has cre- ated breakthroughs in natural language processing applications, such as machine translation (Artetxe et al., 2018; Lample et al., 2018). Those have been mostly based on a bootstrapping approach, which consists in iteratively alternating between two rep- resentations, and optimizing a reconstruction loss. Beyond machine translation, other applications in- clude Question-Answering (Lewis et al., 2019) and parsing (Drozdov et al., 2019). While similar ideas have been applied as well for video summariza- tion (Yuan et al., 2019), such a bootstrapping ap- proach seems less suited for summarization, be- cause of the inherent information loss when going from the full text to the summarized one. Exist- ing unsupervised approaches for summarization therefore relied mostly on extractive graph-based systems (Mihalcea and Tarau, 2004). Only recently have there been proposals for unsupervised abstrac- tive summarization, using auto-encoders (Chu and Liu, 2019; Braˇzinskas et al., 2020). However, these set-ups are complex and require a combination of loss functions (Chu and Liu, 2019) or hierarchical latent variables (Braˇzinskas et al., 2020) to ensure that the generated summaries remain on-topic. In this paper, we investigate a self-supervised ap- proach for multi-document opinion summarization. In this setting, there are multiple opinions (reviews), one entity (products, venues, movies, etc) and the goal is to extract a short summary of those opin- ions. Our approach is based on self-supervision and does not require any gold summaries. We train a supervised model on examples artificially created by selecting (i) one review that will act as a target summary and (ii) a subset of reviews of the same entity that acts as a document collection. Neural models have a known problem of hal- lucination (Rohrbach et al., 2018), which can be misleading …

Content truncated to 500 words