Top-down Discourse Parsing via Sequence Labelling

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 715–726 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 715 Top-down Discourse Parsing via Sequence Labelling Fajri Koto Jey Han Lau Timothy Baldwin School of Computing and Information Systems The University of Melbourne ffajri@student.unimelb.edu.au, jeyhan.lau@gmail.com, tbaldwin@unimelb.edu.au Abstract We introduce a top-down approach to dis- course parsing that is conceptually simpler than its predecessors (Kobayashi et al., 2020; Zhang et al., 2020). By framing the task as a sequence labelling problem where the goal is to iteratively segment a document into individ- ual discourse units, we are able to eliminate the decoder and reduce the search space for splitting points. We explore both traditional recurrent models and modern pre-trained trans- former models for the task, and additionally in- troduce a novel dynamic oracle for top-down parsing. Based on the Full metric, our pro- posed LSTM model sets a new state-of-the-art for RST parsing.1 1 Introduction Discourse analysis involves the modelling of the structure of text in a document. It provides a sys- tematic way to understand how texts are segmented hierarchically into discourse units, and the relation- ships between them. Unlike syntax parsing which models the relationship of words in a sentence, dis- course parsing operates at the document-level, and aims to explain the flow of writing. Studies have found that discourse parsing is beneficial for down- stream NLP tasks including document-level senti- ment analysis (Bhatia et al., 2015) and abstractive summarization (Koto et al., 2019). Rhetorical Structure Theory (RST; Mann and Thompson (1988)) is one of the most widely used discourse theories in NLP (Hernault et al., 2010; Feng and Hirst, 2014; Ji and Eisenstein, 2014; Li et al., 2016; Yu et al., 2018). RST organizes text spans into a tree, where the leaves represent the basic unit of discourse, known as elementary dis- course units (EDUs). EDUs are typically clauses 1Code and trained models: https://github.com/ fajri91/NeuralRST-TopDown EDU-1 EDU-4 EDU-2 EDU-3 elab EDU-1:�Roy�E.�Parrott,�the�company's�president�and�chief� ��������������operating�officer�since�Sept.�1,�was�named�to�its�board. EDU-2:�The�appointment�increased�the�number�of�directors�to�10, EDU-3:�three�of�whom�are�company�employees. EDU-4:�Simpson�is�an�auto�parts�maker. elab elab Figure 1: An example discourse tree, from the RST Discourse Treebank (elab = elaboration). of a sentence. Non-terminal nodes in the tree rep- resent discourse unit relations. In Figure 1, we present an example RST tree with four EDUs spanning two sentences. In this discourse tree, EDUs are hierarchically connected with arrows and the discourse label elab. The direction of arrows indicates the nuclearity of rela- tions, wherein a “satellite” points to its “nucleus”. The satellite unit is a supporting sentence for the nucleus unit and contains less prominent informa- tion. It is standard practice that the RST tree is trained and evaluated in a right-heavy binarized manner, resulting in three forms of binary nuclear- ity relationships between EDUs: Nucleus–Satellite, Satellite–Nucleus, and Nucleus–Nucleus. In this work, eighteen coarse-grained relations are con- sidered as discourse labels, consistent with earlier work (Yu et al., 2018).2 Work on RST parsing has been dominated by the bottom-up paradigm (Hernault et al., 2010; Feng and Hirst, 2014; Ji …

Content truncated to 500 words