Why-Question Answering using Intra- and Inter-Sentential Causal Relations

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 1733–1743, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics Why-Question Answering using Intra- and Inter-Sentential Causal Relations Jong-Hoon Oh∗ Kentaro Torisawa† Chikara Hashimoto ‡ Motoki Sano§ Stijn De Saeger¶ Kiyonori Ohtake∥ Information Analysis Laboratory Universal Communication Research Institute National Institute of Information and Communications Technology (NICT) {∗ rovellia,† torisawa,‡ ch,§ msano,¶stijn,∥ kiyonori.ohtake}@nict.go.jp Abstract In this paper, we explore the utility of intra- and inter-sentential causal relations between terms or clauses as evidence for answering why-questions. To the best of our knowledge, this is the first work that uses both intra- and inter-sentential causal relations for why-QA. We also propose a method for assessing the appropriate- ness of causal relations as answers to a given question using the semantic orienta- tion of excitation proposed by Hashimoto et al. (2012). By applying these ideas to Japanese why-QA, we improved preci- sion by 4.4% against all the questions in our test set over the current state-of-the- art system for Japanese why-QA. In addi- tion, unlike the state-of-the-art system, our system could achieve very high precision (83.2%) for 25% of all the questions in the test set by restricting its output to the con- fident answers only. 1 Introduction “Why-question answering” (why-QA) is a task to retrieve answers from a given text archive for a why-question, such as “Why are tsunamis gen- erated?” The answers are usually text fragments consisting of one or more sentences. Although much research exists on this task (Girju, 2003; Higashinaka and Isozaki, 2008; Verberne et al., 2008; Verberne et al., 2011; Oh et al., 2012), its performance remains much lower than that of the state-of-the-art factoid QA systems, such as IBM’s Watson (Ferrucci et al., 2010). In this work, we propose a quite straightfor- ward but novel approach for such difficult why- QA task. Consider the sentence A1 in Table 1, which represents the causal relation between the cause, “the ocean’s water mass ..., waves are gen- A1 [Tsunamis that can cause large coastal inundation are generated]effect because [the ocean’s water mass is displaced and, much like throwing a stone into a pond, waves are generated.]cause A2 [Earthquake causes seismic waves which set up the water in motion with a large force.]cause This causes [a tsunami.]effect A3 [Tsunamis]effect are caused by [the sudden dis- placement of huge volumes of water.]cause A4 [Tsunamis weaken as they pass through forests]effect because [the hydraulic resistance of the trees diminish their energy.]cause A5 [Automakers in Japan suspended production for an array of vehicles]effect because [the magnitude 9 earthquake and tsunami hit their country on Friday, March 11, 2011.]cause Table 1: Examples of intra/inter-sentential causal relations. Cause and effect parts of each causal re- lation, marked with [..]cause and [..]effect, are con- nected by the underlined cue phrases for causality, such as because, this causes, and are caused by. erated,” and its effect, “Tsunamis ... are gener- ated.” This is a good answer to the question, “Why are tsunamis …

Content truncated to 500 words