Abstract
SNS (Social Network Service) characterized by Facebook and Twitter has become the next generation paradigm of obtaining data, information and knowledge on the web. The aim of this paper is to recommend relevant expert communities to users on the social network by exploiting the extended object-based thesaurus. It is basically an object- based thesaurus taking the urls of domain experts as its instances. Based on the thesaurus, the recommendation is made by inferring relationships between concepts such as "is super/sub of," "is synonym of," "association of" and "user defined." During the inference, the concepts are matched with set of terms extracted from messages of the SNS users and directed by operators added during the semantic analysis of the messages. For example, given a message "those who have experiences about RIA web application using Eclipse," our system infers the relevant concept "Rich Ajax platform" which uses "Eclipse" among RIA web application platforms. Since the concept includes the urls of the corresponding experts resident in a social network, the experts could be recommended to the users through the social network. The inference for the recommendation is implemented as a query evaluation against the thesauri constructed with OTM (Object -based Ontology/Thesaurus Manager). To be shared and to be easily reused on SNS, the thesauri are transformed into XTM (Xml Topic Maps) by OTM after the assignment of proper expert urls to each concept in the thesaurus. For the assignment, we exploit a conventional ranking algorithm applied to each concept, which analyzes papers, reports and related news of the experts to estimate the grade of their expertise. Once the ranked name list of them is obtained together with the associated email list, they are invited to generate their urls in the experimental SNS. Our inference engine adopts its inference mechanism from object inference proposed in OSEM[1], though in a quite different context. It works on the top of Tomcat 6.0, using XTM 1.0 and jQuery 1.4.2. Ten thousands of concepts including synonyms are constructed in the thesaurus for the inference.