Audio structuring and personalized retrieval using ontologies

The goal of this work is to improve the accuracy (precision and recall) and communication effectiveness of a database system response to a user information request, by utilizing a domain-specific ontology. This ontology is employed along with user profile information, to automatically select and deliver appropriate information units from a database. Specifically, the focus is on multimedia audio databases. We are developing a Personal AudioCast system at the USC Integrated Media Systems Center which creates and delivers to an end-user personalized audio programs constructed from an audio database. Techniques are developed to segment and structure audio information, process user requests to generate custom audio results, create and optimize SQL queries, and present the results to the user.