Semantic Graphs for Generating Deep Questions

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1463–1475 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 1463 Semantic Graphs for Generating Deep Questions Liangming Pan1,2 Yuxi Xie3 Yansong Feng3 Tat-Seng Chua2 Min-Yen Kan2 1NUS Graduate School for Integrative Sciences and Engineering 2School of Computing, National University of Singapore, Singapore 3Wangxuan Institute of Computer Technology, Peking University e0272310@u.nus.edu, {xieyuxi, fengyansong}@pku.edu.cn {dcscts@, kanmy@comp.}nus.edu.sg Abstract This paper proposes the problem of Deep Question Generation (DQG), which aims to generate complex questions that require rea- soning over multiple pieces of information of the input passage. In order to capture the global structure of the document and facil- itate reasoning, we propose a novel frame- work which first constructs a semantic-level graph for the input document and then en- codes the semantic graph by introducing an attention-based GGNN (Att-GGNN). After- wards, we fuse the document-level and graph- level representations to perform joint train- ing of content selection and question decod- ing. On the HotpotQA deep-question cen- tric dataset, our model greatly improves per- formance over questions requiring reasoning over multiple facts, leading to state-of-the- art performance. The code is publicly avail- able at https://github.com/WING-NUS/ SG-Deep-Question-Generation. 1 Introduction Question Generation (QG) systems play a vital role in question answering (QA), dialogue system, and automated tutoring applications – by enriching the training QA corpora, helping chatbots start con- versations with intriguing questions, and automati- cally generating assessment questions, respectively. Existing QG research has typically focused on gen- erating factoid questions relevant to one fact ob- tainable from a single sentence (Duan et al., 2017; Zhao et al., 2018; Kim et al., 2019), as exemplified in Figure 1 a). However, less explored has been the comprehension and reasoning aspects of question- ing, resulting in questions that are shallow and not reflective of the true creative human process. People have the ability to ask deep questions about events, evaluation, opinions, synthesis, or reasons, usually in the form of Why, Why-not, How, Input Paragraph A: Pago Pago International Airport Pago Pago International Airport, also known as Tafuna Airport, is a public airport located 7 miles (11.3 km) southwest of the central business district of Pago Pago, in the village and plains of Tafuna on the island of Tutuila in American Samoa, an unincorporated territory of the United States. Input Paragraph B: Hoonah Airport Hoonah Airport is a state-owned public-use airport located one nautical mile (2 km) southeast of the central business district of Hoonah, Alaska. Question: Are Pago Pago International Airport and Hoonah Airport both on American territory? Answer: Yes Input Sentence: Oxygen is used in cellular respiration and released by photosynthesis, which uses the energy of sunlight to produce oxygen from water. Question: What life process produces oxygen in the presence of light? Answer: Photosynthesis a) Example of Shallow Question Generation b) Example of Deep Question Generation Figure 1: Examples of shallow/deep QG. The evidence needed to generate the question are highlighted. What-if, which requires an in-depth understand- ing of the …

Content truncated to 500 words