I Know You Better: User Profile Aware Personalized Dialogue Generation

• Published in: Advanced Data Mining and Applications Vol. 13088; pp. 192 - 205
• Main Authors: Dong, Wenhan, Feng, Shi, Wang, Daling, Zhang, Yifei
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2022; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Dialogue system; Meta learning; Personalized dialogue; Response generation
• ISBN: 9783030954079; 3030954072
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-95408-6_15

Recently, the response generation for dialogue systems has become a research hotspot both in the academic and business communities. Existing personalized response generation methods mainly stand on the chatbot’s perspective, and focus on improving the conversation consistency according to the chatbot’s traits. However, for building an emotionally intelligent and human-like chatbot, it is essential to consider the user’s profile, such as interests, hobbies, and life experiences, and generate the personalized response from the user-oriented perspective. In this paper, we introduce the user profile aware personalized dialogue generation task. For sparse profile users, we extend Model-Agnostic Meta-Learning (MAML) method to quickly adapt to new profiles by leveraging only a few dialogue samples. Extensive experiments are conducted on a real-world dataset, and the results have validated the superiority of the proposed model over strong baseline methods.