基于主题文本的推断社会关系强度的熵模型

在社交网络中，基于主题的文本信息可以反映出用户在主题上的某种兴趣。通过分析用户间的共同兴趣，以推断用户间的关系强度，可用于好友推荐或者广告推送。本文引入多样性和权重频率2种相互独立的方法，共同推断用户的社会关系强度。模型的参数通过已知数据集的卡茨评分进行学习，实验结果表明模型能够较准确地推断用户的社会关系强度。

An Entr opy Model to Infer Social Strength Based on Texts of Subjects

In social networks services, text information based on subjects can reflect users’ certain interest on the subjects. By analyzing the common interests between users, users’ social strength can be inferred, and it can be used to friend recommending and advertisement pushing. This paper introduces two independent ways: diversity and weighted frequency, to infer social strength between users. The parameters of the models are learned from the Katz score of dataset. The results of experiment show that this model can accurately infer the social strength between users.