Modeling User Reviews Through Bayesian Graph Attention Networks for Recommendation

Recommender systems relieve users from cognitive overloading by predicting preferred items for users. Due to the complexity of interactions between and items, graph neural networks (GNN) use structures effectively model user-item interactions. However, existing GNN approaches have following limitations: 1) user reviews are not adequately modeled in graphs. Therefore, preferences item properties that described lost modeling items; 2) GNNs assume deterministic relations which lack stochastic estimate uncertainties neighbor relations. To mitigate limitations, we build tripartite graphs as nodes connect with items. We variables propose a Bayesian attention network ( i.e. , ContGraph ) accurately predict ratings. incorporates prior knowledge regularize posterior inference weights. Our experimental results show significantly outperforms thirteen state-of-the-art models improves best performing baseline ANR) 5.23% on 25 datasets 5-core version. Moreover, correctly semantics can help express