Text Matching Improves Sequential Recommendation by Reducing Popularity Biases

• DOI: 10.48550/arxiv.2308.14029
• Publication Date: 2023-01-01
• Authors: Zhenghao Liu; Sen Mei; Chenyan Xiong; Xiaohua Li; Yu Shi; Zhiyuan Liu; Yu Gu; Yu Ge
• Keywords: Popularity

This paper proposes Text mAtching based SequenTial rEcommendation model (TASTE), which maps items and users in an embedding space recommends by matching their text representations. TASTE verbalizes user-item interactions using identifiers attributes of items. To better characterize user behaviors, additionally attention sparsity method, enables to longer reducing the self-attention computations during encoding. Our experiments show that outperforms state-of-the-art methods on widely used sequential recommendation datasets. alleviates cold start problem representing long-tail full-text modeling bringing benefits pretrained language models systems. further analyses illustrate significantly improves accuracy popularity bias previous item id returning more appropriate text-relevant satisfy users. All codes are available at https://github.com/OpenMatch/TASTE.