Transition Relation Aware Self-Attention for Session-based Recommendation
Guanghui Zhu, Haojun Hou, Jingfan Chen, Chunfeng Yuan, Yihua Huang · Research Square · 2022
Abstract Session-based recommendation is a challenging problem in real-world scenes, e.g., e-commerce, short video platforms, and music platforms,which aims to predict the next click action based on the anonymous session. Recently, graph neural networks (GNNs) have emerged as the state-of-the-art methods for session-based recommendation. However, we find that there exist some limitations in these methods. One is the important transition relations between items are not fully exploited. Another is the graph structure noise and long-range dependency problems when introducing GNNs in session-based recommendation. Considering these problems, we propose a novel approach based on self-attention mechanism for session-based recommendation, called Transition Relation Aware Self-Attention (TRASA). TRASA encodes the shortest path between items in a session graph through the gated recurrent unit as their transition relations. And uses a self-attention mechanism to explore the potential relations between items, which does not rely on the graph structure and can capture the long-range dependencies intrinsically. Besides, the transition relations are incorporated explicitly when computing the attention scores to keep the transition semantics. Extensive experiments on four real-word datasets demonstrate that TRASA outperforms the existing state-of-the-art methods consistently.