An Extraction Method For Ontology Excerpts

The amount of domain knowledge is growing with the relative research goesdeeper. Hence, ontology as an useful tool to express domain knowledge, tends toexpand infinitely. How to deal with the large scale of ontology becomes a tough issuein the field of ontology engineering nowadays. Some measures like ontologymodularization, ontology slicing and approximate reasoning have been proposed tosolve this problem.Jin introduced the notion of ‘ontology slice’ in2007. However, his method basedon graph doesn’t work efficiently on large scale of the ontology. In this paper, weconsider an ontology as a terminology and redefine the notion of ontology slice,called it ontology excerpt. An ontology excerpt is described as a fixed-size subset ofan ontology that preserves as much as possible the knowledge about the terms in agiven signature as described in the ontology.We propose some extraction techniques based on the theory of ontology excerpt.The idea is to rank axioms according to the similarity between the signature and anaxiom and then take top-k axioms as the excerpt we need. Two techniques to calculatethe similarity between signature and axiom are proposed, one is based on commonsignature and the other one is inspired by the knowledge from the field of informationretrieval.To evaluate these techniques, we propose two measures. One is precision and theother one is gain, measuring the degree of incompleteness of the resulting excerptsusing the notion of logical difference. We provide an experimental evaluation of theextraction techniques using different biomedical ontologies. The results show thatgain could sever as a reliable measure to evaluate the quality of excerpts. The main work of this paper is listed as follows:1. Introduced the notion of ontology excerpt2. Described the idea how to extract ontology excerpts3. Considered different techniques to calculate the similarity between signatureand axiom4. Proposed and implement two measures to evaluate the quality of theontology excerpts5. Did a series of experiments to test the effectiveness of the extractiontechniques and measures.The techniques for ontology excerpts provide another solution for processing oflarge ontology. Besides, it could serve as a theoretical basis for ontology reasoning inthe filed of ontology management and ontology debugging.