An Ontology-Based Similarity between Sets of Concepts

To help sharing knowledge in those contexts where documents and services are annotated with semantic information, such as the Semantic Web, defining and implementing the similarity between sets of concepts belonging to a common ontology may prove very useful. In fact, if both the required and the provided pieces of information (be they textual documents, services, images, or whatever) are annotated with sets of concepts taken from a reference ontology O, the evaluation of how good a piece of information P is, w.r.t. the required one R, may be based on the similarity between the two sets of concepts that describe P and R. One of the first applications of the agent technology, aimed at “reducing work and information overload”, was that of retrieving and filtering information in an automatic way. Thus, the possibility to calculate the semantic distance between two sets of concepts finds a natural application in the agent field, in particular for improving those agents that act as “digital butlers” for their human owners, by exploring the Semantic Web and looking for useful documents and/or services. Unfortunately, the metrics for calculating the semantic distance between two sets of concepts that can be found in the literature, are often very simple and do not meet some requirements that, up to us, make the metric closer to the common sense reasoning. For this reason, we have designed and implemented two new algorithms for computing the similarity between sets of concepts belonging to the same ontology.