Dynamic optimization of semantic annotation relevance

The introduction of semantics in the next generation of the Web, the semantic Web, is strongly based on conceptual description of resources by means of semantic annotations. Effective technologies are therefore required lo correctly map the available syntactic information onto a set of relevant conceptual entities able to model the knowledge domain to which a resource belongs. In attempting to address such issue, we propose an evolutionary optimization of semantic annotation relevance which can improve text-to-concept mapping using information from both the syntactic and the semantic domains. The proposed algorithm leverages relevance information on resource contents, with respect to a subset of a given ontology, and performs several ontology navigation steps for extracting the set of most relevant annotations, in terms of semantic expressiveness. The fitness function of the algorithm is strongly time dependent since the set of annotation to be refined may vary according to user requests, to changes in the domain ontology and is related to the granularity of the annotation set.