基于共同属性和标签共现的标签消歧算法

Tag disambiguation model based on common attribute and tag co-occurrence

为了提高基于标签的个性化推荐技术的准确率,提出了一种基于共同属性和标签共现的标签消歧模型,对已有的基于聚类的标签消歧算法进行改进,针对不同的标签语义问题分别采用不同的方法,缓解了原算法不能识别不同语义的问题。对于多义词语义问题,使用同义词模型进行消歧;对于近义词、同义词语义问题,使用近、同义词模型进行消歧,并将该模型应用于个性化推荐算法。利用公共数据集MovieLens Latest Datasets进行了个性化推荐实验。实验表明,当用户推荐项目数量递增时,推荐算法的准确率和召回率都有提高,能有效消除标签中存在的歧义。

To improve the accuracy of personalized recommendation technology based on the label, disambiguation algorithm was put forward based on the common properties and tag the co-occurrence. The existing label disambiguation method based on clustering was improved. Based on semantic problems for different labels in algorithm, different methods of disambiguation were adopted, makes the algorithm does not recognize different semantic. Synonyms disambiguation algorithm was used to solve the problem of polysemous word semantic disambiguation; For synonym with semantic problems, synonyms and near synonyms disambiguation algorithm was used, and the algorithm was applied in the personalized recommendation algorithm. Public data sets Movielens Latest were used to conduct personalized recommendation experiments, which show disambiguation model can effectively improve the recommendation accuracy.