Publications

User Plays a Role: User-insight Multi-modal Recommendation

Published in Journal 1, 2010

Multi-Modal Recommendation (MMRec) aims to help users explore their potential interested items based on multi-modal information input and has been widely used in e-commerce platforms. Recent works mainly focus on modeling item-side information. However, they ignore the abundant semantic information from the user-information modeling, including age, gender, feedback, etc. Such imbalanced attention to item and user leads to inadequate expressiveness of comprehensive interests. In this paper, we propose a novel User-insight Multi-modal recommendation framework, termed UiM. This framework improves user modeling in three aspects: Firstly, to better explore the primary interests from a large-scale item pool, we propose to construct an enriched user profile to re-distribute attention to users’ historical interactions. Secondly, to further disentangle compact representations from heterogeneous items, we propose to apply multi-interest feature extraction on re-attentioned item features. Moreover, an intrinsic shortage of a trivial recommender system is that it fails to access user feedback for in-place result adjustment. As a solution, we access pseudo feedback beforehand from an intelligent agent, then accordingly perform potential adjustments to recommendation candidates for finer results. Extensive experiments show that our model outperforms state-of-the-art multi-modal recommendation models in three public datasets.

Recommended citation: Jingyu Xu, Zechao Hu, Hao Li, et al. User Plays a Role: User-insight Multi-modal Recommendation. IEEE Transactions on Multimedia. 2024. https://KevinXu-01.github.io/home/files/paper2.pdf

Beyond Preferences: Enriching User Profiles for Effective E-commerce Recommendations

Published in Journal 1, 2009

Recommender systems have become a fundamental service in most E-Commerce platforms. Recently, some efforts to extract multi-interests from users’ historical records have demonstrated superior performance. However, aside from historical records, the user profile contains rich semantic information for interest extraction and inherently regulates the users’ interests. Existing works mainly overlook that a user’s interests have: group influence; multi-level preference; and time relevance. To this end, a novel Enhanced User Profile-based Multi-interest Model (E-UPMiM) for recommendation is proposed to integrate historical records with enhanced user profiles with social relationship information to model users’ multi-interests effectively. By leveraging these multiple data sources, our model aims to provide users with more accurate and personalized recommendations. We propose to extract user preferences with three corresponding components: integrated input with enhanced and updated user profile with social relationship information to meet users’ grouping needs; multi-interest extraction module to obtain complex multiple interest representations; and time-aware ranking module to adjust the order of the recommendation list dynamically. Extensive experiments on three public datasets show that E-UPMiM significantly outperforms state-of-the-art recommendation models. From both qualitative and quantitative perspectives, the results demonstrate improvements in recommendation accuracy, personalization, and robustness to changes in user preferences.

Recommended citation: Jingyu Xu, Zhengwei Yang, Zheng Wang. Beyond Preferences: Enriching User Profiles for Effective E-commerce Recommendations. IEEE Transactions on Computational Social Systems. 2024. https://KevinXu-01.github.io/home/files/paper1.pdf