Large Language Models Know Your Contextual Search Intent: A Prompting Framework for Conversational Search

• DOI: 10.18653/v1/2023.findings-emnlp.86
• Publication Date: 2023-12-10T16:58:19Z
• Authors: Kelong Mao; Zhicheng Dou; Fengran Mo; Jiewen Hou; Haonan Chen; Hongjin Qian
• Keywords: Leverage (statistics); Robustness

Precisely understanding users' contextual search intent has been an important challenge for conversational search. As sessions are much more diverse and long-tailed, existing methods trained on limited data still show unsatisfactory effectiveness robustness to handle real scenarios. Recently, large language models (LLMs) have demonstrated amazing capabilities text generation conversation understanding. In this work, we present a simple yet effective prompting framework, called LLM4CS, leverage LLMs as text-based interpreter help Under explore three generate multiple query rewrites hypothetical responses, propose aggregate them into integrated representation that can robustly represent the user's intent. Extensive automatic evaluations human widely used benchmarks, including CAsT-19, CAsT-20, CAsT-21, demonstrate remarkable performance of our LLM4CS framework compared with even using rewrites. Our findings provide evidence better understand