Deep Semantic-Level Cross-Domain Recommendation Model Based on DSV-CDRM

Abstract

A deep semantic-level cross-domain recommendation model based on DSV-CDRM is proposed to address the problems of existing methods such as single modeling approach. First, review information is converted into word vectors using a TinyBERT pre-trained language model, and then two global deep semantic viewpoint matrices are used in conjunction with a gating mechanism to guide queries. An additional convolutional layer is added on top of the improved text convolution to construct auxiliary documents using similar but non-overlapping user comments. Finally, correlations between deep semantic viewpoints between different domains are learned by constructing a correlation matrix and performing semantic matching. Experiments on the Amazon public dataset demonstrate that the proposed method outperforms existing models in both MAE and MSE, and it can effectively improve the performance of cross-domain recommendation system.