Abstract: The user interests and product features are extracted from comments in traditional recommendation algorithms. However, the expected recommendation performance is not achieved as it is difficult to obtain valid information, caused by the free-form and poor regularity of comments. In the current field of electronic commerce, the comment tag as a new way of comments has been widely used. Compared with comments, the comment tag has the advantages of strong regularity and information density. Thus this paper proposes a recommendation algorithm fusing comment tag which extracts the users’ opinions for the product features and then makes use of them to construct user interests model and product features model. Therefore, the proposed algorithm can recommend the products with well-reviews on specific features which users are interested in. Compared with traditional algorithms, the experimental results show that the proposed algorithm can effectively improve the user coverage and the recommendation accuracy.

Key words: comment tag, product feature, recommendation algorithm

摘要： 传统的推荐算法大都从评论中挖掘用户兴趣或产品特征，然而由于评论形式自由，规则性差，导致从评论中获取有效信息较困难，推荐结果不理想。在电子商务等领域，评论标签作为一种新的评论方式已经被广泛使用。与评论相比，评论标签具有规则性强，信息密度大等特点，因此提出了一种融合评论标签的推荐算法。该算法从评论标签中挖掘用户对产品特征的观点，并利用其构建用户兴趣模型和产品特征模型，然后向用户推荐在他们感兴趣的特征上有较高评价的产品。与传统推荐算法进行对比，实验结果表明，融合评论标签的算法能有效地提高用户的覆盖率，并提升推荐算法的准确性。

关键词: 评论标签, 产品特征, 推荐算法