基于评论挖掘和用户偏好学习的评分预测协同过滤

Collaborative filtering based on opinion mining and user preference learning

提出了一种新的方法挖掘评论中的文字信息,将评论对象被用户关注的层面发掘出来并评分,根据这些层面的分数以及用户过往的评分数据学习出用户的偏好,最后根据用户的偏好预测其他待评分对象的分数并产生推荐。实验结果表明,提出的方法在预测准确度方面较传统方法有一定程度的提高。

This paper proposed a new method that made use of text comments,discovered the aspects user matters,then predicated the rating for each aspect. It learned user＇s preference from the past rating regard to the aspect ratings. Using the user preference and the aspect ratings,it generated the predicated ratings. Experiments show that the proposed method is more accuracy than the classic ones.