Person Correction Based On Temporal Sematic Path Similarity In Heterogeneous Networks

Social networking has been one of the most popular fields in contemporary. In the process of Social networking building, the phenomena of relations clutter become more obvious due to the diversity of data sources, the expanding of data scale, the imperfectness of transection information, the discrepancy of information mode and structure as well as former name used. For this reason, it is difficult for us to recognize the uniqueness of person in social network. The traditional analysis measures of social networking based on common networks cannot solve this problem. Based on heterogeneous networks and sematic networks, this paper present a person uniqueness measure based on temporal sematic path similarity computing method and verification of same person merging based on structure error calculation.After surveying the traditional common network model and heterogeneous network model, we analyzed the limitation of common network model in large-scale social network analysis and then proposed person uniqueness measure based on temporal sematic path similarity in heterogeneous social networks. This method proposed the concept of temporal sematic networks based upon the characteristics of sematic networks, which computes temporal sematic path similarity between two nodes. We implemented the measure of person uniqueness in social networks based upon similarity threshold and unique person filtering method. All the experimental results showed that our method can mesure person uniqueness precisely in large-scale social networks and verified the effectiveness of the method.To the phenomenon of highly similar person in social networks, we proposed the method of networks structure error calculation in social networks. This method calculates the structure error of node pairs by the changes of person node degree before and after merging. We can use this method to filtrate the node that have the same relation structure. Then, we devised the method of same person merging in social networks to implement person correction. After analyzing an academic network with many types of academic activity information, we verified the effectiveness of our method.