An efficient method to measure the semantic similarity of ontologies

With the recent availability of large number of bioinformatics data sources, query from such databases and rigorous annotation of experimental results often use semantic frameworks in the form of an ontology. With the growing access to heterogeneous and independent data repositories, determining the semantic similarity or difference of two ontologies is critical in information retrieval, information integration and semantic web services. The purpose of this paper is to propose a new sense refinement algorithm to construct a refined sense set (RSS) for an ontology so that the senses (synonym words) in this refined sense set represent the semantic meanings of the terms used by this ontology.

A new concept of a semantic set is introduced that combines the refined sense set of ontology with the relationship edges connecting the terms in this ontology to represent the semantics of this ontology. With the semantic sets, measuring the semantic similarity or difference of two ontologies is simplified as comparing the commonality or difference of two sets.

The experimental studies show that the proposed method of measuring the semantic similarity or difference of two ontologies is efficient and accurate; comparisons with existing methods show the efficacy of using the new method.

The concepts introduced in this paper will improve automation of bioinformatics databases to serve queries based on heterogeneous ontologies.