What is it about?
It is about using LLMs to dynamically analyze the personalization requirement for each conversation turn, then combine non-personalized and personalized search results accordingly.
Featured Image
Why is it important?
It is a timely topic since current commercial chatbots (ChatGPT, Gemini, Claude, etc.) are adding such textual user profiles in their product, aiming at memorizing some important facts about the users and thus using them to optimize future user experience. Another reason is that such user-centirc approaches can help satisfy the special information needs by minority groups, which are usually overlooked by traditional search engins.
Perspectives
This is my first publication specialized in Information Retrieval. Through this project, I see some fundamental limitations of current IR models. They still cannot understant complex information need. On the other hand, I observed that unlike other NLP tasks, IR can not benifit that much from the emergence of LLMs. This means IR is probably one of the most difficult problem to solve.
Yuchen Hui
Universite de Montreal
Read the Original
This page is a summary of: Towards Adaptive Personalized Conversational Information Retrieval, November 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3746252.3761255.
You can read the full text:
Contributors
The following have contributed to this page
