Improving Implicit Discourse Relation Classification by Modeling Inter-dependencies of Discourse Units in a Paragraph

Proceedings of NAACL-HLT 2018, pages 141–151 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Improving Implicit Discourse Relation Classification by Modeling Inter-dependencies of Discourse Units in a Paragraph Zeyu Dai, Ruihong Huang Department of Computer Science and Engineering Texas A&M University {jzdaizeyu, huangrh}@tamu.edu Abstract We argue that semantic meanings of a sentence or clause can not be interpreted independently from the rest of a paragraph, or independently from all discourse relations and the overall paragraph-level discourse structure. With the goal of improving implicit discourse relation classification, we introduce a paragraph-level neural networks that model inter-dependencies between discourse units as well as discourse relation continuity and patterns, and predict a sequence of discourse relations in a paragraph. Experimental results show that our model out- performs the previous state-of-the-art systems on the benchmark corpus of PDTB. 1 Introduction PDTB-style discourse relations, mostly defined between two adjacent text spans (i.e., discourse units, either clauses or sentences), specify how two discourse units are logically connected (e.g., causal, contrast). Recognizing discourse relations is one crucial step in discourse analysis and can be beneficial for many downstream NLP applications such as information extraction, machine transla- tion and natural language generation. Commonly, explicit discourse relations were distinguished from implicit ones, depending on whether a discourse connective (e.g., “because” and “after”) appears between two discourse units (Prasad et al., 2008a). While explicit dis- course relation detection can be framed as a dis- course connective disambiguation problem (Pitler and Nenkova, 2009; Lin et al., 2014) and has achieved reasonable performance (F1 score > 90%), implicit discourse relations have no dis- course connective and are especially difficult to identify (Lin et al., 2009, 2014; Xue et al., 2015). To fill the gap, implicit discourse relation pre- diction has drawn significant research interest re- cently and progress has been made (Chen et al., 2016; Liu and Li, 2016) by modeling composi- tional meanings of two discourse units and ex- ploiting word interactions between discourse units using neural tensor networks or attention mecha- nisms in neural nets. However, most of existing approaches ignore wider paragraph-level contexts beyond the two discourse units that are examined for predicting a discourse relation in between. To further improve implicit discourse relation prediction, we aim to improve discourse unit rep- resentations by positioning a discourse unit (DU) in its wider context of a paragraph. The key obser- vation is that semantic meaning of a DU can not be interpreted independently from the rest of the paragraph that contains it, or independently from the overall paragraph-level discourse structure that involve the DU. Considering the following para- graph with four discourse relations, one relation between each two adjacent DUs: (1): [The Butler, Wis., manufacturer went pub- lic at $15.75 a share in August 1987,]DU1 and (Explicit-Expansion) [Mr. Sim’s goal then was a $29 per-share price by 1992.]DU2 (Implicit-Expansion) [Strong earnings growth helped achieve that price far ahead of sched- ule, in August 1988.]DU3 (Implicit-Comparison) [The stock has since softened, trading around $25 a share …

Content truncated to 500 words