MEGA RST Discourse Treebanks with Structure and Nuclearity from Scalable Distant Sentiment Supervision

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 7442–7457, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 7442 MEGA RST Discourse Treebanks with Structure and Nuclearity from Scalable Distant Sentiment Supervision Patrick Huber and Giuseppe Carenini Department of Computer Science University of British Columbia Vancouver, BC, Canada, V6T 1Z4 {huberpat, carenini}@cs.ubc.ca Abstract The lack of large and diverse discourse tree- banks hinders the application of data-driven approaches, such as deep-learning, to RST- style discourse parsing. In this work, we present a novel scalable methodology to auto- matically generate discourse treebanks using distant supervision from sentiment-annotated datasets, creating and publishing MEGA-DT, a new large-scale discourse-annotated corpus. Our approach generates discourse trees in- corporating structure and nuclearity for doc- uments of arbitrary length by relying on an efficient heuristic beam-search strategy, ex- tended with a stochastic component. Experi- ments on multiple datasets indicate that a dis- course parser trained on our MEGA-DT tree- bank delivers promising inter-domain perfor- mance gains when compared to parsers trained on human-annotated discourse corpora. 1 Introduction Discourse parsing is an important Natural Lan- guage Processing (NLP) task, aiming to uncover the hidden structure underlying coherent docu- ments, as described by theories of discourse like Rhetorical Structure Theory (RST) (Mann and Thompson, 1988) or PDTB (Prasad et al., 2008). Not only has discourse parsing been shown to en- hance key downstream tasks, such as text classifica- tion (Ji and Smith, 2017), summarization (Gerani et al., 2014) and sentiment analysis (Bhatia et al., 2015; Nejat et al., 2017; Hogenboom et al., 2015), but it also appears to complement contextual em- beddings, like BERT (Devlin et al., 2018), in tasks where discourse information is critical, such as ar- gumentation analysis (Chakrabarty et al., 2019). Traditionally, RST-style discourse parsing builds a complete, hierarchical constituency tree for a doc- ument (Morey et al., 2018), where leaf nodes are clause-like sentence fragments, called elementary- discourse-units (EDUs), while internal tree nodes are labelled with discourse relations (e.g., Evi- dence, Contrast). In addition, each node is given a nuclearity attribute, which encodes the importance of the node in its local context. A key limitation for further research in RST-style discourse parsing is the scarcity of training data. Only a few human annotated discourse treebanks exist, each only containing a few hundred docu- ments. Although our recent efforts using distant supervision from sentiment to generate large-scale discourse treebanks have already partly addressed this dire situation (Huber and Carenini, 2019), the previously proposed solution is still limited in: (i) Scope, by only building the RST constituency struc- ture without nuclearity and relation labels; and (ii) Applicability, by relying on a non-scalable CKY so- lution, which cannot be applied to many real-world datasets with especially long documents. In this work, we propose a significant extension to this line of research by introducing a scalable so- lution for documents of arbitrary length and further moving beyond just predicting the tree-structure by incorporating the nuclearity attribute, oftentimes critical in informing downstream tasks (Marcu, 2000; Ji …

Content truncated to 500 words