To solve the problem of sparsity of comment information in the cross-domain recommendation method based on user comment text,a cross-domain hybrid recommendation method based on comment text and professional rating was proposed.The attention mechanism and gate control mechanism were used to extract aspect features from comment text,and a global cross-domain aspect correlation matrix was constructed for matching.The rating information of review text was combined to emphasize the importance of different users'ratings of items,the user expertise was introduced to refine the user's ratings of items after aggregation.Experimental results show that the proposed method can improve the accuracy of recommendation.
关键词
深度学习/冷启动/评分推荐/跨域推荐/专业度/注意力机制/数据稀疏
Key words
deep learning/cold-start/rating recommendations/cross-domain recommendation/expertise/attention mechanism/data sparsity
维普
万方数据