Discourse-Aware Unsupervised Summarization for Long Scientific Documents

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 1089–1102 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 1089 Discourse-Aware Unsupervised Summarization of Long Scientific Documents Yue Dong∗ MILA/McGill University Montreal, QC, Canada yue.dong2 @mail.mcgill.ca Andrei Mircea∗ MILA/McGill University Montreal, QC, Canada andrei.romascanu @mail.mcgill.ca Jackie C. K. Cheung MILA/McGill University Montreal, QC, Canada jcheung @cs.mcgill.ca Abstract We propose an unsupervised graph-based rank- ing model for extractive summarization of long scientific documents. Our method as- sumes a two-level hierarchical graph represen- tation of the source document, and exploits asymmetrical positional cues to determine sen- tence importance. Results on the PubMed and arXiv datasets show that our approach1 outperforms strong unsupervised baselines by wide margins in automatic metrics and human evaluation. In addition, it achieves perfor- mance comparable to many state-of-the-art su- pervised approaches which are trained on hun- dreds of thousands of examples. These results suggest that patterns in the discourse structure are a strong signal for determining importance in scientific articles. 1 Introduction Single document summarization aims at shorten- ing a text and preserving the most important ideas of the source document. While abstractive strate- gies generate summaries with novel words, extrac- tive strategies select sentences from the source to form a summary (Nenkova et al., 2011). Despite recent advances in abstractive summarization, ex- tractive models are still attractive in cases where faithfully preserving the original text is the priority. For example, legal arguments can hinge on the ex- act wording of a contract (Farzindar and Lapalme, 2004), and ensuring the factual correctness of a summary can be critical in the health or scientific domains, which is a known weakness of current abstractive methods (Kry´sci´nski et al., 2019). Supervised neural-based models have been the dominant paradigm in recent extractive systems, at least for short news summarization (Nallapati et al., ∗Equal contribution. 1Link to our code: https://github.com/ mirandrom/HipoRank. Introduction anxiety affects quality of life in those living with parkinson’s disease (pd) more so than overall cognitive status, motor deficits, apathy, and depression. Introduction although anxiety and depression are often re- lated and coexist in pd patients, recent research suggests that anxiety rather than depression is the most prominent and prevalent mood disor- der in pd. Related Work furthermore, since previous work, albeit lim- ited, has focused on the influence of symptom laterality on anxiety and cognition, we also explored this relationship . Methodology this study is the first to directly compare cog- nition between pd patients with and without anxiety. Result the findings confirmed our hypothesis that anxiety negatively influences attentional set- shifting and working memory in pd. Result moreover, anxiety has been suggested to play a key role in freezing of gait (fog), which is also related to attentional set-shifting. Future work s. future research should examine the link between anxiety, set-shifting, and fog, in order to determine whether treating anxiety might be a potential therapy for improving fog. Table 1: Example of a PubMed article’s summary pro- duced by our …

Content truncated to 500 words