Multi-Task Learning for Conversational Question Answering over a Large-Scale Knowledge Base

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 2442–2451, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 2442 Multi-Task Learning for Conversational Question Answering over a Large-Scale Knowledge Base Tao Shen1∗, Xiubo Geng2, Tao Qin2, Daya Guo3, Duyu Tang2, Nan Duan2, Guodong Long1 and Daxin Jiang2 1Centre for AI, School of Computer Science, FEIT, University of Technology Sydney 2Microsoft, Beijing, China 3The School of Data and Computer Science, Sun Yat-sen University tao.shen@student.uts.edu.au, guodong.long@uts.edu.au {xiubo.geng,taoqin,dutang,nanduan,djiang}@microsoft.com guody5@mail2.sysu.edu.cn Abstract We consider the problem of conversational question answering over a large-scale knowl- edge base. To handle huge entity vocabulary of a large-scale knowledge base, recent neu- ral semantic parsing based approaches usu- ally decompose the task into several subtasks and then solve them sequentially, which leads to following issues: 1) errors in earlier sub- tasks will be propagated and negatively affect downstream ones; and 2) each subtask cannot naturally share supervision signals with oth- ers. To tackle these issues, we propose an in- novative multi-task learning framework where a pointer-equipped semantic parsing model is designed to resolve coreference in conver- sations, and naturally empower joint learn- ing with a novel type-aware entity detection model. The proposed framework thus en- ables shared supervisions and alleviates the ef- fect of error propagation. Experiments on a large-scale conversational question answering dataset containing 1.6M question answering pairs over 12.8M entities show that the pro- posed framework improves overall F1 score from 67% to 79% compared with previous state-of-the-art work. 1 Introduction Recent decades have seen the development of AI- driven personal assistants (e.g., Siri, Alexa, Cor- tana, and Google Now) that often need to an- swer factorial questions. Meanwhile, large-scale knowledge base (KB) like DBPedia (Auer et al., 2007) or Freebase (Bollacker et al., 2008) has been built to store world’s facts in a structure database, which is used to support open-domain question an- swering (QA) in those assistants. Neural semantic parsing based approach (Jia and Liang, 2016; Reddy et al., 2014; Dong and ∗ Work done while the author was an intern at Microsoft, Beijing, China. Lapata, 2016; Liang et al., 2016; Dong and Lap- ata, 2018; Guo et al., 2018) is gaining rising atten- tion for knowledge-based question answer (KB- QA) in recent years since it does not rely on hand- crafted features and is easy to adapt across do- mains. Traditional approaches usually retrieve an- swers from a small KB (e.g., small table) (Jia and Liang, 2016; Xiao et al., 2016) and are difficult to handle large-scale KBs. Many recent neural se- mantic parsing based approaches for KB-QA take a stepwise framework to handle this issue. For example, Liang et al. (2016), Dong and Lapata (2016), and Guo et al. (2018) first use an entity linking system to find entities in a question, and then learn a model to map the question to logi- cal form based on that. Dong and Lapata (2018) decompose the semantic parsing …

Content truncated to 500 words