Bridging the User-side Knowledge Gap in Knowledge-aware Recommendations with Large Language Models

TL;DR

This work tackles the underexplored problem of structuring user side knowledge for knowledge aware recommendations by leveraging large language models to infer explicit user interests and constructing a Collaborative Interest Knowledge Graph (CIKG) that fuses user side, item side, and collaborative signals. It introduces CIKGRec, a two stage framework consisting of LLM based CIKG construction and a GNN based recommendation model that incorporates a user interest reconstruction module to mitigate LLM noise and a cross domain contrastive learning module to enable knowledge transfer. The model is trained with a joint objective that combines a ranking loss, a noise robust reconstruction loss and a cross domain contrastive loss, along with a KG translation loss to preserve entity semantics. Experiments on three real world datasets show state of the art performance, especially for users with sparse interactions, validating the effectiveness of the proposed noise suppression and cross domain transfer mechanisms in knowledge aware recommendations.

Abstract

In recent years, knowledge graphs have been integrated into recommender systems as item-side auxiliary information, enhancing recommendation accuracy. However, constructing and integrating structural user-side knowledge remains a significant challenge due to the improper granularity and inherent scarcity of user-side features. Recent advancements in Large Language Models (LLMs) offer the potential to bridge this gap by leveraging their human behavior understanding and extensive real-world knowledge. Nevertheless, integrating LLM-generated information into recommender systems presents challenges, including the risk of noisy information and the need for additional knowledge transfer. In this paper, we propose an LLM-based user-side knowledge inference method alongside a carefully designed recommendation framework to address these challenges. Our approach employs LLMs to infer user interests based on historical behaviors, integrating this user-side information with item-side and collaborative data to construct a hybrid structure: the Collaborative Interest Knowledge Graph (CIKG). Furthermore, we propose a CIKG-based recommendation framework that includes a user interest reconstruction module and a cross-domain contrastive learning module to mitigate potential noise and facilitate knowledge transfer. We conduct extensive experiments on three real-world datasets to validate the effectiveness of our method. Our approach achieves state-of-the-art performance compared to competitive baselines, particularly for users with sparse interactions.

Figures (7)

• Figure 1: A toy example of the progress from traditional methods to the collaborative interest knowledge graph. $u_1$ is the target user to provide recommendation for. The green circle denote the important users and items discovered by user-side higher-order knowledge but is overlooked by traditional methods.
• Figure 2: The overall architecture of our method. The left half illustrates the LLM-based user-side knowledge construction, and the right half shows the CIKG-based recommendation.
• Figure 3: Performance comparisons under different sparsity user groups.
• Figure 4: A real case in Book-Crossing dataset.
• Figure 5: Hyperparameter sensitive experiments on Book-Crossing.