Diversity Matters: User-Centric Multi-Interest Learning for Conversational Movie Recommendation
Abstract: Diversity plays a crucial role in Recommender Systems (RSs) as it ensures a wide range of recommended items, providing users with access to new and varied options. Without diversity, users often encounter repetitive content, limiting their exposure to novel choices. While significant efforts have been dedicated to enhancing recommendation diversification in static offline scenarios, relatively less attention has been given to online Conversational Recommender Systems (CRSs). However, the lack of recommendation diversity in CRSs will increasingly exacerbate over time due to the dynamic user-system feedback loop, resulting in challenges such as the Matthew effect, filter bubbles, and echo chambers. To address these issues, we propose an innovative end-to-end CRS paradigm called User-Centric Multi-Interest Learning for Conversational Movie Recommendation (CoMoRec), which aims to learn user interests from multiple perspectives to enhance result diversity as users engage in natural language conversations for movie recommendations. Firstly, CoMoRec automatically models various facets of user interests, including context-based, graph-based, and review-based interests, to explore a wide range of user intentions and preferences. Then, it leverages these multi-aspect user interests to accurately predict personalized and diverse movie recommendations and generate fluent and informative responses during conversations. Through extensive experiments conducted on two publicly available CRS-based movie datasets, our proposed CoMoRec achieves a new state-of-the-art performance and outperforms all the compared baselines in terms of improving recommendation diversity in the CRS.
Relevance To Conference: The work "Diversity Matters: User-Centric Multi-Interest Learning for Conversational Movie Recommendation" makes a notable contribution to the field of multimedia processing by addressing the challenge of improving recommendation diversity in conversational movie recommendation systems. In the realm of multimedia processing, effectively leveraging multiple modalities, such as text, images, and user interactions, is crucial for enhancing recommendation quality. This work contributes by proposing a user-centric multi-interest learning framework that tackles the issue of recommendation diversity specifically in conversational movie recommendations. By capturing and modeling diverse user interests expressed during conversations, the proposed framework enables the generation of personalized and diverse movie recommendations. This is achieved by incorporating various multimodal signals, including textual dialogues, user ratings, and visual movie posters, to comprehensively understand user preferences. The authors introduce an attention mechanism that dynamically weights the importance of different modalities, thereby facilitating adaptive learning from user interactions and incorporating relevant modalities during recommendation generation. Overall, this work contributes to the advancement of multimedia processing by extending traditional recommendation systems to the conversational context and emphasizing the importance of diversity in recommendations. By integrating multimodal data and considering user-centric modeling, it enhances recommendation accuracy while providing users with more engaging and varied movie recommendations. This research demonstrates the significance of addressing recommendation diversity in multimedia applications, improving the overall recommendation experience for users.
Primary Subject Area: [Engagement] Multimedia Search and Recommendation
Submission Number: 879
Loading