Interactive Information Need Prediction with Intent and Context

TL;DR

Interactive Information Need Prediction introduces a framework where users provide a pre-search context and an optional partial intent to predict their formal information need. The study evaluates generation (to form a full question) and retrieval (to fetch an answer) using adapted Inquisitive and MS MARCO datasets, across encoder–decoder, prompting LLMs, and retrieval models. Key findings show that partial intent mitigates the distraction from large contexts and that retrieval (especially cross-encoder) often yields strong results, while generation with a tuned encoder–decoder also performs well. The work demonstrates feasibility for real-world interactive search systems and provides code and data to enable further research.

Abstract

The ability to predict a user's information need would have wide-ranging implications, from saving time and effort to mitigating vocabulary gaps. We study how to interactively predict a user's information need by letting them select a pre-search context (e.g., a paragraph, sentence, or singe word) and specify an optional partial search intent (e.g., "how", "why", "applications", etc.). We examine how various generative language models can explicitly make this prediction by generating a question as well as how retrieval models can implicitly make this prediction by retrieving an answer. We find that this prediction process is possible in many cases and that user-provided partial search intent can help mitigate large pre-search contexts. We conclude that this framework is promising and suitable for real-world applications.

Figures (2)

• Figure 1: Demonstrating information need prediction via selected pre-search context and specified partial search intent.
• Figure 2: The trade-off between selected pre-search context and specified search intent.