CATCH: A Controllable Theme Detection Framework with Contextualized Clustering and Hierarchical Generation

2025-12-25

AI Summary

This work addresses the challenge of cross-dialogue topic detection in user-centric dialogue systems without predefined schemas—particularly difficult due to weak topic representations in short, sparse utterances, poor inter-dialogue consistency, and the absence of user-specific preference modeling. We propose the first unified framework integrating context-aware representation learning, preference-guided clustering, and hierarchical topic generation. Built upon an 8B-parameter LLM, it incorporates context-enhanced encoding, joint clustering with personalized feedback, and hierarchical decoding to achieve robust topic semantic modeling and controllable label generation. Evaluated on the DSTC-12 multi-domain customer-service benchmark, our method significantly outperforms state-of-the-art approaches in both topic clustering purity and label readability. To our knowledge, it is the first unsupervised method to simultaneously ensure inter-turn consistency and user-level preference alignment.

Technology Category

• Natural Language Processing: Conversational AI/Dialog Systems
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Learning Preferences or Rankings

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search

Abstract

Theme detection is a fundamental task in user-centric dialogue systems, aiming to identify the latent topic of each utterance without relying on predefined schemas. Unlike intent induction, which operates within fixed label spaces, theme detection requires cross-dialogue consistency and alignment with personalized user preferences, posing significant challenges. Existing methods often struggle with sparse, short utterances for accurate topic representation and fail to capture user-level thematic preferences across dialogues. To address these challenges, we propose CATCH (Controllable Theme Detection with Contextualized Clustering and Hierarchical Generation), a unified framework that integrates three core components: (1) context-aware topic representation, which enriches utterance-level semantics using surrounding topic segments; (2) preference-guided topic clustering, which jointly models semantic proximity and personalized feedback to align themes across dialogue; and (3) a hierarchical theme generation mechanism designed to suppress noise and produce robust, coherent topic labels. Experiments on a multi-domain customer dialogue benchmark (DSTC-12) demonstrate the effectiveness of CATCH with 8B LLM in both theme clustering and topic generation quality.

Problem

• Detects latent topics in user dialogues without predefined schemas
• Ensures cross-dialogue consistency and aligns with user preferences
• Handles sparse, short utterances and captures user-level thematic preferences

Innovation

• Context-aware topic representation enriches utterance semantics
• Preference-guided topic clustering aligns themes across dialogues
• Hierarchical theme generation suppresses noise for robust labels