Ontology Extraction by Collaborative Tagging with Social Networking
Masahiro Hamasaki, Yutaka Matsuo, Takuichi Nishimura, Hideaki Takeda · 2008
This paper proposes integration of a social network with col-laborative tagging for ontology extraction. Tripartite mod-els of emergent ontologies based on three dimensions (i.e. users, tags, and instances) have been proposed by several researchers, but we integrate another important dimension: user-user relations, such as the friend relation in social net-working services and a knows relation in Friend-Of-A-Friend (FOAF) documents. Because a lightweight ontology is a minimum commitment shared within a community, who com-municates with whom is an important source of information that can be used to improve the emergent ontology. We also discuss the advanced model in where each concept in each community is considered different (and called p-concept), and show the possibility of using this model to resolve the polysemy/hononymy problem. Two case studies using our algorithms are shown: we analyze tagging and social net-working data from an academic conference support system POLYPHONET and also from an advanced social system called Blue Dot. We evaluate the extracted ontologies for information recommendation, and show that our algorithm works better than others.