GTM: A Generative Triple-wise Model for Conversational Question Generation

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 3495–3506 August 1–6, 2021. ©2021 Association for Computational Linguistics 3495 GTM: A Generative Triple-Wise Model for Conversational Question Generation Lei Shen1,2 Fandong Meng3 Jinchao Zhang3 Yang Feng1,2∗ Jie Zhou3 1Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China 2University of Chinese Academy of Sciences, Beijing, China 3Pattern Recognition Center, WeChat AI, Tencent Inc, China shenlei17z@ict.ac.cn, {fandongmeng,dayerzhang}@tencent.com fengyang@ict.ac.cn, withtomzhou@tencent.com Abstract Generating some appealing questions in open- domain conversations is an effective way to improve human-machine interactions and lead the topic to a broader or deeper direction. To avoid dull or deviated questions, some researchers tried to utilize answer, the “fu- ture” information, to guide question genera- tion. However, they separate a post-question- answer (PQA) triple into two parts: post- question (PQ) and question-answer (QA) pairs, which may hurt the overall coherence. Besides, the QA relationship is modeled as a one-to-one mapping that is not reasonable in open-domain conversations. To tackle these problems, we propose a generative triple-wise model with hierarchical variations for open-domain con- versational question generation (CQG). Latent variables in three hierarchies are used to rep- resent the shared background of a triple and one-to-many semantic mappings in both PQ and QA pairs. Experimental results on a large- scale CQG dataset show that our method sig- nificantly improves the quality of questions in terms of fluency, coherence and diversity over competitive baselines. 1 Introduction Questioning in open-domain dialogue systems is indispensable since a good system should have the ability to well interact with users by not only re- sponding but also asking (Li et al., 2017). Besides, raising questions is a proactive way to guide users to go deeper and further into conversations (Yu et al., 2016). Therefore, the ultimate goal of open- domain conversational question generation (CQG) is to enhance the interactiveness and maintain the continuity of a conversation (Wang et al., 2018). Joint work with Pattern Recognition Center, WeChat AI, Tencent Inc, China. ∗Yang Feng is the corresponding author. Post: I ate out with my friends this evening. Question Candidates: Q1.1: Which restaurant did you go? Q1.2: Where did you eat? Q2.1: What food did you eat? Q2.2: Did you eat something special? Q3: What do you mean? Q4: How about drinking together? Answer Candidates: A1: We went to an Insta-famous cafeteria. A2: We ate steak and pasta. Table 1: An example of CQG task which is talking about a person’s eating activity. There are one-to-many mappings in both PQ and QA pairs. The content of each meaningful and relevant question (Q1.1 to Q2.2) is decided by its post and answer. Q3 (dull) and Q4 (deviated) are generated given only the post. CQG differs fundamentally from traditional ques- tion generation (TQG) (Zhou et al., 2019; Kim et al., 2019; Li et al., 2019) that generates a ques- tion given a sentence/paragraph/passage and a spec- ified answer within it. While in CQG, an answer always …

Content truncated to 500 words