Title: NBQ: Next-Best-Question for Dynamic Profiling

Original: arXiv:2606.00809v1 Announce Type: new Abstract: Many real-world conversational settings for knowledge discovery, including podcasts, hiring screens, and marketplaces, require a purpose-driven understanding of a person. We study the Next-Best-Question (NBQ) problem: at each turn, an interviewer should ask the question with the highest expected information gain given what has already been learned and the conversation goal. We propose NBQ, a plug-and-play framework that seeds a diverse pool of candidate questions, maintains a compact and continuously updated user state, adaptively selects the next question within a turn budget, and distills the resulting free-form dialogue into a structured vector-based user profile. As a demanding application, we instantiate NBQ for reciprocal matchmaking, where compatibility must be mutual and each person is modeled by both self-description and counterpart-preference representations. To support large-scale matching, we further introduce QuickMatch, an efficient retrieval layer that recasts reciprocal matching from quadratic pairwise scoring to approximate vector search. Experiments show that NBQ improves user profiling quality by up to 13.6% and 14.0% in AC@T and AR@T, respectively, while QuickMatch accelerates retrieval by up to 22.9x with recall up to 0.989.

Rewritten:

Abstract: Purposeful comprehension of individuals is essential in numerous practical dialogue scenarios aimed at knowledge extraction, such as podcast interviews, recruitment screenings, and commercial marketplaces. This paper investigates the Next-Best-Question (NBQ) challenge, which posits that an interviewer should select the query yielding the maximum expected information gain at every step, conditioned on prior knowledge and the specific objective of the interaction. To address this, we introduce NBQ, a modular framework designed to generate a varied set of potential questions, sustain a lean and dynamically refreshed user profile, and strategically choose subsequent inquiries within a defined turn limit. This process converts unstructured conversational data into a structured, vector-based user representation. We demonstrate the framework’s utility in the complex domain of reciprocal matchmaking, where mutual compatibility is required, and individuals are characterized by both their self-descriptions and their preferences for partners. To facilitate scalable matching operations, we also present QuickMatch, a high-efficiency retrieval mechanism that transforms the computationally intensive quadratic pairwise scoring process into a fast approximate vector search. Our experimental results indicate that NBQ enhances user profiling accuracy by as much as 13.6% in AC@T and 14.0% in AR@T. Furthermore, QuickMatch boosts retrieval speed by up to 22.9 times while maintaining a recall rate of up to 0.989.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC