Abstract: Many personalized resource recommendation methods based on social tagging ignore the different affects on recommendation of users’ long and short interests and polysemous tag for improving precision.In order to solve these problems，the metrics are designed for making a clear distinction between user’s long and short interests—— user’s tag preference weight and user’s resource preference weight.Based on the above，a novel hybrid recommend method has been proposed，which equilibrates both strength of content-based and collaborative filtering recommendation.The proposed method can eliminate influence of polysemous tag by adding the calculation factor of tagging the same resource tab vectors similarity.The experimental results show that this method for personalized resource recommendation in the social tagging system outperforms other recommendation algorithms.

Key words: social tagging system, personalized information recommendation, recommendation algorithm, user profile, tag

摘要： 目前许多基于社化化标注的个性化资源推荐方法均忽视了用户长短期兴趣和多义标签问题对推荐的不同影响，为此，设计区分用户长短期兴趣的指标——用户的标签偏好权重和资源偏好权重；在此基础上，提出一种结合基于内容和基于协同过滤方法优点的混合推荐方法，通过加入标注相同资源的标签向量相似度计算因子，来减小多义标签对推荐结果的影响。实验表明，将该方法引入社会化标注系统资源个性化推荐算法中，能提高推荐精度。

关键词: 社会化标注系统, 个性化信息推荐, 推荐算法, 用户模型, 标签