基于评论主题分析的评分预测方法研究

发布时间：2018-03-13 20:22

本文选题：推荐系统　切入点：评分预测　出处：《中文信息学报》2017年02期 　论文类型：期刊论文

【摘要】：推荐系统(recommender system)广泛应用于电子商务网站。目前流行的基于协同过滤的推荐算法利用用户的历史评分来预测用户对物品的喜好程度。随着互联网的发展,如今的电子商务网站越来越注重与用户的交互,于是产生了大量的用户生成内容(user generated content),如评论、地理位置、好友关系等。相对评分来说,用户对物品的评论从用户或者物品的各个角度具体表达了用户的观点。利用这些信息更有助于挖掘用户的喜好。该文提出一种基于词向量的方法挖掘用户评论信息,并结合协同过滤的方法设计新的推荐算法,来改善评分预测的效果。实验结果表明,该算法较大程度上提高了评分预测精度。
[Abstract]:Recommendation system is widely used in e-commerce websites. Currently popular recommendation algorithms based on collaborative filtering use the user's history score to predict the user's preference for goods. Today's e-commerce sites are increasingly focused on interacting with users, resulting in a lot of user-generated generated content, such as reviews, geographic locations, friendships, etc. Users' comments on items express their views from various angles of users or items. Using these information is more helpful for mining users' preferences. This paper proposes a word-vector-based approach to mining user comment information. A new recommendation algorithm is designed with collaborative filtering to improve the effect of score prediction. The experimental results show that the algorithm improves the accuracy of score prediction to a large extent.
【作者单位】：苏州大学计算机科学与技术学院;
【分类号】：TP391.3

【相似文献】