Reproduce “How to Leverage Personal Textual Knowledge for Personalized Conversational Information Retrieval” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Fengran Mo, Longxiang Zhao, Kaiyu Huang, Yue Dong, Degen Huang, Jian-Yun Nie
This research developed and evaluated strategies for integrating personal textual knowledge into personalized conversational information retrieval systems using large language models. The study found that selecting personal knowledge based on its direct impact on retrieval effectiveness, rather than human intuition, yielded superior performance, and that LLM-aided query reformulation, guided by a few in-context examples, improved retrieval results on relevant conversational turns.
Key insights
- Human intuition about PTKB relevance does not consistently correlate with its actual impact on retrieval performance; an impact-driven automatic annotation method is more effective.
- Large Language Models can effectively perform complex tasks like PTKB selection and personalized query reformulation in data-scarce environments, especially with targeted in-context learning.
- Personalization is not universally beneficial; applying it indiscriminately or with irrelevant PTKB information can be detrimental to retrieval effectiveness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.