基于评论文本和融入专业度评分的跨域混合推荐

Cross-domain hybrid recommendation based on comment text and professional rating

为解决基于用户评论文本的跨域推荐方法产生的评论信息稀疏性问题,提出一种基于评论文本和融入专业度评分的跨域混合推荐方法。采用注意力机制和门控机制对评论文本进行方面特征抽取,构建全局跨域方面相关矩阵进行匹配,结合评论文本中的评分信息,生成一个跨域粗矩阵以降低原始评分矩阵的稀疏度。为强调不同用户对项目评分的重要性,引入用户专业度细化聚合后用户对项目的评分。实验结果表明,该方法可以提高推荐的准确性。

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.