A Framework for Automatic Ontology Generation from Autonomous Web Applications

Ontologies capture the structure, relationships, semantics and other essential meta information of an application. This thesis describes a framework to automate application interoperability by using dynamically generated ontologies. We propose a set of techniques to extract ontologies from data accessible on the Web in the form of semi-structured HTML pages. Ontologies retrieved from similar applications are matched together to create a general ontology describing the application domain. Information retrieval and graph matching techniques are used to match and measure the usefulness of the ontologies created. Matching algorithms are combined together to produce global ontologies based on local ontologies inherently present in Web applications. We present a system called OntoBuilder that allows users to drive the ontology creation process using a userriendly and intuitive interface. We also present experiments for a well-known case of study: car-rental applications. We successfully achieve 90% accuracy on ontology extraction and 70% accuracy for ontology matching