A Unified Linear-Time Framework for Sentence-Level Discourse Parsing

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4190–4200 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 4190 A Unified Linear-Time Framework for Sentence-Level Discourse Parsing Xiang Lin∗¶, Shafiq Joty∗¶§, Prathyusha Jwalapuram¶ and M Saiful Bari¶ ¶Nanyang Technological University, Singapore §Salesforce Research Asia, Singapore {linx0057@e., srjoty@, jwal0001@e., bari0001@e.}ntu.edu.sg Abstract We propose an efficient neural framework for sentence-level discourse analysis in ac- cordance with Rhetorical Structure Theory (RST). Our framework comprises a discourse segmenter to identify the elementary discourse units (EDU) in a text, and a discourse parser that constructs a discourse tree in a top-down fashion. Both the segmenter and the parser are based on Pointer Networks and operate in lin- ear time. Our segmenter yields an F1 score of 95.4, and our parser achieves an F1 score of 81.7 on the aggregated labeled (relation) met- ric, surpassing previous approaches by a good margin and approaching human agreement on both tasks (98.3 and 83.0 F1). 1 Introduction Coherence analysis of a text is a fundamental task in Natural Language Processing that can benefit many downstream applications. Rhetorical Struc- ture Theory or RST (Mann and Thompson, 1988) is one of the most influential theories of text coher- ence. According to RST, a text is represented by a hierarchical structure known as a Discourse Tree (DT). As exemplified in Figure 1, the leaves of a DT correspond to contiguous atomic text spans called Elementary Discourse Units (EDUs). The adjacent EDUs and larger units are recursively connected by certain coherence relations (e.g., AT- TRIBUTION, EXPLANATION). The discourse units connected by a relation are further categorized based on their relative importance: NUCLEUS refers to the core part(s), while SATELLITE refers to the peripheral one. Coherence analysis in RST involves two subtasks: (i) breaking the text into a sequence of EDUs, referred to as Discourse Seg- mentation, and (ii) linking the EDUs into a DT, referred to as Discourse Parsing. ∗*equal contribution CONDITION ATTRIBUTION e1 S e2 N N TEMPORAL e3 N e4 N S [The Treasury also said]e1 [noncompetitive tenders will be considered timely]e2 [if postmarked no later than Sunday, Oct.29,]e3 [and received no later than tomorrow.]e4 Figure 1: An example discourse tree with four EDUs. In this paper we consider sentence-level coher- ence analysis, which involves discourse segmen- tation and sentence-level parsing. For example, consider the DT in figure 1 for the sentence “The Treasury also said noncompetitive tenders will be considered timely if postmarked no later than Sun- day, Oct.29, and received no later than tomorrow.”, which has four EDUs as shown below the tree. Such sentence-level discourse annotations have been shown to be beneficial for a number of ap- plications including machine translation (Guzm´an et al., 2014) and sentence compression (Sporleder and Lapata, 2005). Furthermore, sentence-level analysis is considered to be a crucial step towards full text-level analysis. For example, automatic discourse segmentation has been shown to be the main source of inaccuracies in discourse parsing (Soricut and Marcu, 2003; Joty …

Content truncated to 500 words