Evaluating semantic similarity with a new method of path analysis in RDF using genetic algorithms

Abstract

This paper presents a novel method of evaluating semantic similarity by means of path analysis in RDF databases. Similarity is calculated by assignining each property (predicate in RDF terms) a weight, which is found using a genetic optimization algorithm. Presented method exhibits an advatage over existing methods, because of its flexibility and the fact that no prior knowledge of a particular database is necessary. This paper also presents an exemplary application of the method - recommendation engine. Proposed method is applied to a well known problem - music recommendation based on DBPedia. Results obtained in the experiment positively verify its advanntages and usefulness.