Finding semantic similarity in social networks by using preference management and profiling

Today, internet and social networks are involved more and more in our daily lives as a means of communication and information tool. Volume of data generated in social networks has reached critical levels and continues to increase. Storage and management of such huge data to make smart decisions creates effects to ease daily lives of people. The accuracy of these effects is possible by expressing the personal preferences of people using Internet and social networks. Personal preferences are captured by observing and recording of the interactions between social network users and social network items and later used to create personal services for individuals. Semantic web aims to represent data in a common representation to make it interoperable for computers. Inside semantic web applications, data is represented inside ontologies. Expressing and managing personal preferences inside ontologies enables personal data stored in different social networks to be used in smart decisions. People always wants to be with someone they felt similar to. Being in an environment that includes differences enforce people to adapt these differences to blend in. The similarity between people is to expose and store these differences. The concept of similarity is used to give better services to people among different fields. Finding similar people and creating the more accurate recommendation for these people results as better personalized services. However, personal similarity can be achieved by modeling and representing of individual's behaviors. On the contrary, similarity is achieved by comparison of these representations and capturing similar behaviors. In this thesis, a new method to find semantic similarities among people is introduced by modeling the behaviors of people in social networks inside preferences. In order to apply this method in social networks, an ontology-based entity semantification method that provides matching of personal preferences with entities in ontologies has been defined. The success of this method is tested with similar works in literature and conclude that our method gave the best results. As a result of semantification, an integrated ontology dataset was created in which personal preferences are represented as entities inside ontologies and stored in personal profiles. Social semantic similarity method is applied to selected samples of this dataset. The success of the method was tested by simple similarity obtained by using non-semantic data and page similarity obtained by using semi-structural data.