Inducing Ontologies from Folksonomies using Natural Language Understanding

Social bookmarking systems enable users to tag and organize documents according to their own subjective preferences. The collection of tags or folksonomy describes common resources of interest to the user community. The implicit meaning associated with the terms in a tag enables the discovery of other relevant documents and users with shared interests. Folksonomy lacks the hierarchy structure of taxonomy that facilitates better information access or the semantics of an ontology that yields to machine reasoning. The project proposes to make explicit the latent information structure in folksonomies by using natural language processing and automatic ontology generation from text. Phase I effort focuses on (1) the linguistic tools required to capture tag semantics from the tag text, social annotation and the textual content of tagged documents, (2) knowledge classification tools to organize the extracted semantics in an ontology, and (2) the candidate applications that can demonstrate the use of tag semantics. Different information access and discovery applications can use the underlying ontology structure to enable users browse the tag space, and mine new information and associations in social bookmarked data.