CQG: A Simple and Effective Controlled Generation Framework for Multi-hop Question Generation

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 6896 - 6906 May 22-27, 2022 c⃝2022 Association for Computational Linguistics CQG: A Simple and Effective Controlled Generation Framework for Multi-hop Question Generation Zichu Fei1,2, Qi Zhang1,2∗, Tao Gui3∗, Di Liang4, Sirui Wang4, Wei Wu4, Xuanjing Huang1,2,3 School of Computer Science, Fudan Unviersity1 Shanghai Key Laboratory of Intelligent Information Processing, Shanghai, China2 Institute of Modern Languages and Linguistics, Fudan University, Shanghai, China3 Meituan Inc., Beijing, China4 {zcfei19, qz, tgui,xjhuang}@fudan.edu.cn {liangdi04, wangsirui}@meituan.com, wuwei19850318@gmail.com Abstract Multi-hop question generation focuses on generating complex questions that require reasoning over multiple pieces of information of the input passage. Current models with state-of-the-art performance have been able to generate the correct questions corresponding to the answers. However, most models can not ensure the complexity of generated questions, so they may generate shallow questions that can be answered without multi-hop reasoning. To address this challenge, we propose the CQG, which is a simple and effective controlled framework. CQG employs a simple method to generate the multi-hop questions that contain key entities in multi-hop reasoning chains, which ensure the complexity and quality of the questions. In addition, we introduce a novel controlled Transformer-based decoder to guarantee that key entities appear in the questions. Experiment results show that our model greatly improves performance, which also outperforms the state-of-the-art model about 25% by 5 BLEU points on HotpotQA 1. 1 Introduction Question generation (QG) aims to endow machines with the ability to ask relevant and to-the-point questions about a document. QG plays a vital role in question answering (QA), dialogue systems, and automated tutoring applications: – by enriching the training QA corpora (Tang et al., 2017; Yuan et al., 2017), helping chatbots start conversations with intriguing questions (Mostafazadeh et al., 2016), and automatically generating assessment questions (Heilman and Smith, 2010), respectively. Most prior research on QG has focused on shallow factoid-based questions where answering the question simply by extracting the span of the text from a single input document (Zhou et al., ∗ Corresponding authors. 1Our code and models are publicly available at https://github.com/sion-zcfei/CQG Figure 1: An example that the uncontrolled question generation model may generate the correct but shallow questions. In this example, the model ignores the important entity Western European in paragraph A and then generate a shallow question without multi-hop reasoning chains. 2018; Zhao et al., 2018; Kim et al., 2019; Fei et al., 2021). Recently, motivated by building the NLP systems that are capable of understanding and reasoning (Kaushik and Lipton, 2018; Sinha et al., 2019), there is an increasing interest in developing systems that are capable of more complex multi-hop question generation, where answering the questions requires reasoning over multiple documents (Pan et al., 2020; Sachan et al., 2020; Xie et al., 2020; Yu et al., 2020; Su et al., 2020). Compared with shallow QG, there are two 6896 challenges for multi-hop QG (MQG). At first, generating multi-hop questions requires the model to understand the relationship between disjointed …

Content truncated to 500 words