Empowering Retrieval-based Conversational Recommendation with Contrasting User Preferences

2025-03-27

AI Summary

Existing conversational recommender systems typically employ a monolithic user representation, making it difficult to distinguish opposing affective intents—such as “like” versus “dislike”—leading to explicit preference ambiguity and suboptimal recommendation performance. To address this, we propose a preference-decoupled retrieval-based conversational recommendation framework. First, we design a contrastive preference expansion mechanism that leverages large language models to infer implicit preferences. Second, we introduce explicit preference-decoupled representation learning to separately model positive and negative user intents. Our approach integrates preference-aware representation, contrastive learning, and retrieval-augmented recommendation. Evaluated on three benchmark datasets, our method achieves up to a 99.72% improvement in Recall@10 over prior state-of-the-art methods. It is the first to enable interpretable, fine-grained decoupling of opposing intents while delivering highly effective recommendations.

Technology Category

Data Mining & Knowledge Management: Conversational Systems for Recommendation & RetrievalMachine Learning: Learning Preferences or RankingsKnowledge Representation and Reasoning: Preferences

Application Category

User Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingSemantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Conversational recommender systems (CRSs) are designed to suggest the target item that the user is likely to prefer through multi-turn conversations. Recent studies stress that capturing sentiments in user conversations improves recommendation accuracy. However, they employ a single user representation, which may fail to distinguish between contrasting user intentions, such as likes and dislikes, potentially leading to suboptimal performance. To this end, we propose a novel conversational recommender model, called COntrasting user pReference expAnsion and Learning (CORAL). Firstly, CORAL extracts the user's hidden preferences through contrasting preference expansion using the reasoning capacity of the LLMs. Based on the potential preference, CORAL explicitly differentiates the contrasting preferences and leverages them into the recommendation process via preference-aware learning. Extensive experiments show that CORAL significantly outperforms existing methods in three benchmark datasets, improving up to 99.72% in Recall@10. The code and datasets are available at https://github.com/kookeej/CORAL

Problem

• Distinguishing contrasting user intentions in recommendations
• Improving accuracy by capturing user likes and dislikes
• Enhancing conversational recommender systems with preference-aware learning

Innovation

• Contrasting preference expansion using LLMs
• Differentiates and leverages contrasting preferences
• Preference-aware learning for recommendation process