Abstract:Knowledge graph (KG)-based recommendation methods effectively alleviate the data sparsity and cold-start problems in collaborative filtering. Among these methods, neighborhood-based methods are the mainstream methods. However, these methods ignore some meta-information about the items, specifically, the diversity of item information (e.g., texts) and feature interaction between neighboring nodes. In this paper, we propose a Bilinear Knowledge-aware Graph Neural Network Fusing Text Information (BKGNN-TI), which can model both knowledge graph information and text information. In particular, the information in knowledge graph contains not only the existing high-order connectivity but also feature interactions between neighboring nodes at the same level in KG. First, we construct the information propagation layer using the bilinear collector and linear collector. Feature interactions between neighboring nodes and the high-order connectivity are collected in the information propagation layer to generate the item knowledge representations. The bilinear collector emphasizes the importance of second-order feature interaction between neighboring nodes in the KG. Then, texts are also introduced when computing the item representations, which can help further infer user interests. We choose objective program titles and introductions as text information to avoid the influence of subjective factors. BKGNN-TI designs an ALBERT-based sequence encoder to encode texts by the structure of ALBERT+Bi-LSTM+Attention, thus enriching the feature representations of the items. In the experiments, we utilize two language datasets, i.e., the English public dataset Movielens-20M and the Chinese dataset IPTV constructed by ourselves. The results demonstrate that our BKGNN-TI outperforms baselines, indicating that our BKGNN-TI is a generalization for both Chinese and English datasets.

BIKAGCN: Knowledge-Aware Recommendations Under Bi-layer Graph Convolutional Networks

To see further: Knowledge graph-aware deep graph convolutional network for recommender systems

BI-GCN: Bilateral Interactive Graph Convolutional Network for Recommendation

Knowledge-enhanced graph convolutional network for recommendation

Knowledge Graph Convolutional Networks for Recommender Systems

KLGCN: Knowledge graph-aware Light Graph Convolutional Network for recommender systems

LighterKGCN: A Recommender System Model Based on Bi-layer Graph Convolutional Networks

KGIE: Knowledge graph convolutional network for recommender system with interactive embedding

BKGNN-TI: A Bilinear Knowledge-Aware Graph Neural Network Fusing Text Information for Recommendation

BiInfGCN: Bilateral Information Augmentation of Graph Convolutional Networks for Recommendation

A Recommendation Strategy Integrating Higher-Order Feature Interactions with Knowledge Graphs

KGCNA: Knowledge Graph Collaborative Neighbor Awareness Network for Recommendation

Cultural Resource Recommender System with Knowledge Graph Convolutional Network.

A novel Knowledge Graph recommendation algorithm based on Graph Convolutional Network

Multi-modal Knowledge Graph Convolutional Network for Recommendation

Knowledge-aware fine-grained attention networks with refined knowledge graph embedding for personalized recommendation

Multi-Behavior Enhanced Heterogeneous Graph Convolutional Networks Recommendation Algorithm based on Feature-Interaction

KGCFRec: Improving Collaborative Filtering Recommendation with Knowledge Graph

Double End Knowledge Graph Convolutional Networks for Recommender Systems

Attentive Knowledge-aware Graph Convolutional Networks with Collaborative Guidance for Personalized Recommendation

Personalize d knowle dge-aware recommendation with collaborative and attentive graph convolutional networks