Self-Supervised learning for Conversational Recommendation

doi:10.1016/j.ipm.2022.103067

	Self-Supervised learning for Conversational Recommendation
	Li, Shuokai 1,2; Xie, Ruobing 3; Zhu, Yongchun 1,2; Zhuang, Fuzhen 4,5; Tang, Zhenwei 6; Zhao, Wayne Xin 7,8; He, Qing 1,2
	2022-11-01
发表期刊	INFORMATION PROCESSING & MANAGEMENT
ISSN	0306-4573
卷号	59 期号:6 页码:19
摘要	Conversational recommender system (CRS) aims to model user preference through interactive conversations. Although there are some works, they still have two drawbacks: (1) they rely on large amounts of training data and suffer from data sparsity problem; and (2) they do not fully leverage different types of knowledge extracted from dialogues. To address these issues in CRS, we explore the intrinsic correlations of different types of knowledge by self-supervised learning, and propose the model SSCR, which stands for Self-Supervised learning for Conversational Recommendation. The main idea is to jointly consider both the semantic and structural knowledge via three self-supervision signals in both recommendation and dialogue modules. First, we carefully design two auxiliary self-supervised objectives: token-level task and sentence-level task, to explore the semantic knowledge. Then, we extract the structural knowledge based on external knowledge graphs from user mentioned entities. Finally, we model the inter-information between the semantic and structural knowledge with the advantages of contrastive learning. As existing similarity functions fail to achieve this goal, we propose a novel similarity function based on negative log-likelihood loss. Comprehensive experimental results on two real-world CRS datasets (including both English and Chinese with about 10,000 dialogues) show the superiority of our proposed method. Concretely, in recommendation, SSCR gets an improvement about 5% similar to 15% compared with state-of-the-art baselines on hit rate, mean reciprocal rank and normalized discounted cumulative gain. In dialogue generation, SSCR outperforms baselines on both automatic evaluations (distinct n-gram, BLEU and perplexity) and human evaluations (fluency and informativeness).
关键词	Conversational recommender system Self-supervised learning Knowledge
DOI	10.1016/j.ipm.2022.103067
收录类别	SCI
语种	英语
资助项目	National Natural Science Foundation of China[61976204] ; National Natural Science Foundation of China[U1811461] ; National Natural Science Foundation of China[62176014] ; National Natural Science Foundation of China[U1836206]
WOS研究方向	Computer Science ; Information Science & Library Science
WOS类目	Computer Science, Information Systems ; Information Science & Library Science
WOS记录号	WOS:000861188400006
出版者	ELSEVIER SCI LTD
引用统计	被引频次：12[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/19825
专题	中国科学院计算技术研究所期刊论文
通讯作者	Zhuang, Fuzhen; He, Qing
作者单位	1.Chinese Acad Sci, Inst Comp Technol, Key Lab Intelligent Informat Proc, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 3.Tencent, WeChat Search Applicat Dept, Shenzhen, Peoples R China 4.Beihang Univ, Inst Artificial Intelligence, Beijing 100191, Peoples R China 5.Beihang Univ, Sch Comp Sci, SKLSDE, Beijing 100191, Peoples R China 6.Univ Toronto, Toronto, ON, Canada 7.Renmin Univ, Gaoling Sch Artificial Intelligence, Beijing, Peoples R China 8.Beijing Acad Artificial Intelligence, Beijing Key Lab Big Data Management & Anal Methods, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Li, Shuokai,Xie, Ruobing,Zhu, Yongchun,et al. Self-Supervised learning for Conversational Recommendation[J]. INFORMATION PROCESSING & MANAGEMENT,2022,59(6):19.
APA	Li, Shuokai.,Xie, Ruobing.,Zhu, Yongchun.,Zhuang, Fuzhen.,Tang, Zhenwei.,...&He, Qing.(2022).Self-Supervised learning for Conversational Recommendation.INFORMATION PROCESSING & MANAGEMENT,59(6),19.
MLA	Li, Shuokai,et al."Self-Supervised learning for Conversational Recommendation".INFORMATION PROCESSING & MANAGEMENT 59.6(2022):19.