Generating Self-Contained and Summary-Centric Question Answer Pairs via Differentiable Reward Imitation Learning

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 5103–5135 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 5103 Generating Self-Contained and Summary-Centric Question Answer Pairs via Differentiable Reward Imitation Learning Li Zhou Kevin Small Yong Zhang Sandeep Atluri Amazon Alexa {lizhouml,smakevin,yonzhn,satluri}@amazon.com Abstract Motivated by suggested question generation in conversational news recommendation sys- tems, we propose a model for generating question-answer pairs (QA pairs) with self- contained, summary-centric questions and length-constrained, article-summarizing an- swers. We begin by collecting a new dataset of news articles with questions as titles and pair- ing them with summaries of varying length. This dataset is used to learn a QA pair gener- ation model producing summaries as answers that balance brevity with sufficiency jointly with their corresponding questions. We then reinforce the QA pair generation process with a differentiable reward function to mitigate ex- posure bias, a common problem in natural lan- guage generation. Both automatic metrics and human evaluation demonstrate these QA pairs successfully capture the central gists of the ar- ticles and achieve high answer accuracy.1 1 Introduction Automatic generation of question-answer pairs (QA pairs) is a widely studied problem, primar- ily used to improve the performance of question answering systems via data augmentation (Alberti et al., 2019; Shakeri et al., 2020). However, ques- tion generation has also recently garnered interest in the context of conversational agents, where sug- gested questions (SQs) (i.e., You can also ask...) have emerged as a promising approach to drive multi-turn dialogues by educating customers about the agent capabilities and guiding users along dia- logue trajectories with more engaging content (Yin et al., 2020; Nouri et al., 2020). As an example, consider a news chatbot engaged in a dialogue regarding COVID-19 vaccine devel- opments producing the SQ {Q: How effective is the Pfizer-BioNTech vaccine?} paired with the answer 1Code and Data will be made available at https:// github.com/amazon-research/SC2QA-DRIL {A: Pfizer/BioNTech vaccine is around 91% effec- tive at preventing COVID-19, according to updated trial data. Experts fear new variants of COVID-19 from South Africa and Brazil may be resistant to existing vaccines and treatment.} Firstly, SQs of this form mitigates the user burden regarding the necessity of both deep subject knowledge to ask good questions and awareness of the agent ques- tion answering capabilities to expect good answers. Secondly, the agent can look-ahead when select- ing SQs to bias toward confidently correct answers and content expected to lead to further follow-up questions and general system engagement. Targeting the SQ problem in news chatbot sce- narios (e.g., (Laban et al., 2020)), this work exam- ines QA pair generation corresponding to a news article summary paired with a self-contained ques- tion. Table 1 shows an example of the task. SQs based on these summary-centric QA pairs act as implicit article recommendations, complementing SQs focusing on passage-level extracted answers or factoid information. QA pairs generated for this purpose must satisfy several criteria includ- ing: (1) questions are self-contained (i.e., users need not read the corresponding articles nor re- …

Content truncated to 500 words