Semantic-enhanced Contrastive Learning for Session-based Recommendation

Zhicheng Liu, Yulong Wang, Tongcun Liu, Lei Zhang, Wei Li, Jianxin Liao, Ding He

Abstract: Highlights•We highlight the problem of limited sample space and false negative samples in contrastive learning for session-based recommendation.•A momentum-updated queue is introduced to expand the sample range.•A semantic-enhanced mechanism is designed to filter out false negative samples and assign different weights to the true negative samples.•Experiments on three real-world datasets verify our model in terms of both performance and efficiency.