Semantic Profile Of Online Social Relationship With Interactive Language

The profile of online social relationship is fundamental in social network analysis and many social computing studies such as community discovery,influence analysis and link predication.A comprehensive profile should describe the nature of a relationship on two levels: properties and semantics.In current studies,because of the convenience in representation and measurement,the properties are more widely considered than the semantics in social relationship profile.However,the semantics can also be very helpful to understand the social relationship.Current studies mainly characterize the semantics of relationships with positive/negative signs or fixed categories which are not sufficient to reveal the specific semantics of a certain relationship.In sociolinguistics,interactive language in communication is assumed to be highly related to the semantics of social relationship.Inspired by this,we propose to semantically model the connotation of social relationship with interactive language between individuals.We connect the properties and content topics of the interactive language with the semantics of the social relationship and then propose to profile the general and detailed semantics of social relationship in three styles,using the high-level features,semantic vectors and topic words of interactive language.We choose two different datasets to do the experiments: English Enron emails and Chinese Sina microblogs.Experimental results illustrate that the proposed profiles can help us to distinguish,compare and understand the semantics of social relationships.