Evolving from semantic network notions, ontologies have become a cornerstone in supporting interoperability and facilitating knowledge management and configuration. With the availability of useable tools, the number of ontologies increased rapidly, leading to more and more complex knowledge bases containing redundant, useless, or not very well structured information with regard to their usage. This has a negative impact on the efficiency of the semantic-based systems, especially at the reasoner and the storage level. Within this project we propose an ontology refactoring technique that aims to improve the efficiency of semantics-based systems by adapting the system’s internal description of ontologies. The changes proposed by the refactoring process will operate only on the system’s internal representation of an ontology and will not be reflected into original ontology structure. We propose the fundamental research for finding an optimal system’s internal description of an ontology with regard to its usage, by analyzing, specifying and implementing a prototype for ontology refactoring in an evolutionary framework. In the following we present the main steps planed in our project:
- Ontology classification – We will investigate if the techniques used by the existing tools for the creation and the maintenance of ontologies reveal important characteristics about the internal structure of different kinds of knowledge bases. We will use this information to create a statistical classification theory for ontologies that will be also considered in order to assess the system’s performance based on ontology descriptions created by the ontology refactoring process.
- Ontology refactoring measures – The refactoring process will be guided by a metric that comprises evaluations at both the storage and the inference level. All these measures are defined based on the statistical information gathered from the ontology structure.
- Ontological remodeling operators – For refactoring an ontology description, some changes must be applied to its internal representation, in accordance with the semantics induced by the original ontology. Thus we propose to develop a set of remodeling operators grouped in three categories:
- deletion functions – remove resource definitions based on their interrelationships with other resources;
- replacement functions – absorb the definition of one concept by another concept semantically interrelated;
- exploration functions – explore new ontological configurations by changing the associations between classes and relations or by refining the concepts based on their class extensions.
- Evolutionary framework for ontology refactoring - Optimizing the internal structure of an ontology description by applying remodeling operators requires a search technique to find a sequence of operators that transform the ontology internal representation. Due to the capabilities of evolutive methods to model complex optimization problems, we intend to design an evolutive framework for ontology refactoring. Important issues such as genetic representation, objective function, evolutive operators have to be carefully designed and analyzed.