Towards Knowledge-Based Recommender Dialog System

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 1803–1813, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 1803 Towards Knowledge-Based Recommender Dialog System Qibin Chen1, Junyang Lin2, Yichang Zhang2, Ming Ding1, Yukuo Cen1, Hongxia Yang2, Jie Tang1 1Department of Computer Science and Technology, Tsinghua University 2DAMO Academy, Alibaba Group {cqb19,dm18,cyk18}@mails.tsinghua.edu.cn {junyang.ljy,yichang.zyc,yang.yhx}@alibaba-inc.com jietang@tsinghua.edu.cn Abstract In this paper, we propose a novel end-to- end framework called KBRD, which stands for Knowledge-Based Recommender Dialog System. It integrates the recommender sys- tem and the dialog generation system. The dialog system can enhance the performance of the recommendation system by introducing knowledge-grounded information about users’ preferences, and the recommender system can improve that of the dialog generation system by providing recommendation-aware vocabu- lary bias. Experimental results demonstrate that our proposed model has significant advan- tages over the baselines in both the evaluation of dialog generation and recommendation. A series of analyses show that the two systems can bring mutual benefits to each other, and the introduced knowledge contributes to both their performances.1 1 Introduction Dialog in e-commerce has great commercial po- tential. In conventional recommender systems, personalized recommendation is highly based on the previous actions of users, including searching, clicking and purchasing. These actions can be re- garded as users’ feedbacks that reflect users’ in- terest. However, due to its implicitness, such feed- back can only reflect a part of users’ interest, caus- ing inaccuracy in recommendation. Another infor- mation source about user preferences is the dialog between users and services. In such dialog, users often provide more information about their pref- erences. They often ask for tips or recommenda- tion in the dialog. In this process, services can guide them to speak out their interests in order to solve users’ problems and meet their require- ments. Compared with the implicit feedback, the 1Code will be available at https://github.com/ qibinc/KBRD. USER: Hello! RECOMMENDER: What kind of movies do you like? USER: I am looking for a movie recom- mendation. When I was younger I really enjoyed the A Nightmare on Elm Street (1984). BASELINE: Have you seen It (2017)? OURS: I like horror movies too! Have your seen Halloween (1978) ? HUMAN: Oh, you like scary movies? recently watched Happy DeathI Day (2017). Table 1: An example of the outputs of recommender dialog systems. The model recommends items (italic) while maintaining the dialog with the user. Compared with the baseline, our dialog system gives a more di- verse and consistent response. feedback from the dialog is more explicit and more related to users’ preferences. Therefore, a recom- mender dialog system possesses high commercial potential. A recommender dialog system can be regarded as a combination of a recommender system and a dialog system. A dialog system should re- spond to users’ utterances with informative natu- ral language expressions, and a recommender sys- tem should provide high-quality recommendation based on the content of users’ utterances. We demonstrate an example …

Content truncated to 500 words