Exploiting Session Information in BERT-based Session-aware Sequential Recommendation
This addresses the need for efficient session-aware recommendation in online services, though it appears incremental as it builds on existing BERT-based models with specific modifications.
The paper tackled the problem of incorporating session information in sequential recommendation systems, proposing three lightweight methods (session tokens, session segment embeddings, and time-aware self-attention) that improved recommendation performance while minimizing additional parameters.
In recommendation systems, utilizing the user interaction history as sequential information has resulted in great performance improvement. However, in many online services, user interactions are commonly grouped by sessions that presumably share preferences, which requires a different approach from ordinary sequence representation techniques. To this end, sequence representation models with a hierarchical structure or various viewpoints have been developed but with a rather complex network structure. In this paper, we propose three methods to improve recommendation performance by exploiting session information while minimizing additional parameters in a BERT-based sequential recommendation model: using session tokens, adding session segment embeddings, and a time-aware self-attention. We demonstrate the feasibility of the proposed methods through experiments on widely used recommendation datasets.