CIIR Talk Series: "Conversation Personalization: From Data Collection to Response Generation"
Title: Conversation Personalization: From Data Collection to Response Generation
Abstract: Personalization is paramount for conversational search and recommendation. Despite the tremendous progress in Large Language Models (LLMs), LLM-based conversational agents tend to deliver generalized recommendations that fail to capture the nuanced interests of individual users. A major hindrance in developing a personal conversational system is the unavailability of large-scale conversational datasets that reflect actual user preferences and interactions. In this talk, I present a method to collect large-scale, multi-session, multi-domain, and human-written personal conversations. Further, I discuss methods for extracting personal information, generating user profiles, and providing personalized responses. By utilizing user preferences in a semi-structured format, LLMs can better recall previously disclosed preferences when generating personalized recommendations. Conversation personalization is around the corner, and LLMs can help users to shape the future of personalization.
Bio: Faegheh Hasibi is an assistant professor at the Data Science group of the Institute of Computing and Information Sciences (iCIS) at Radboud University. Her research interests span different areas in Information Retrieval and Natural Language Processing, with a particular focus on conversational AI and knowledge-grounded search systems. She received her PhD from Norwegian University of Science and Technology, titled “Semantic Search with Knowledge bases.” She has served as the chair of the ICTIR steering committee, as the general chair of ICTIR 2021, and as a program co-chair of SIGIR-AP 2024. Hasibi is the recipient of the best (short) paper honorable mention awards from ICTIR ’16 and SIGIR ’17.
To obtain the passcode for this series, please see the event advertisement on the seminars email list or reach out to ataubman [at] cs.umass.edu (Alex Taubman).
For any questions about this event with the Center for Intelligent Information Retrieval, please contact jean [at] cs.umass.edu (subject: CIIR%20Talk%20Series) (Jean Joyce).