Query approximation by semantic similarity in GeoPQL

This paper proposes a method for query approximation in Geographical Information Systems. In our approach, queries are expressed by the Geographical Pictorial Query Language (GeoPQL), and query approximation is performed using WordNet, a lexical database for the English language available on the Internet. Due to the focus on geographical context, we address WordNet partition taxonomies. If a concept contained in a query has no match in the database, the query is approximated using the immediate superconcepts and subconcepts in the WordNet taxonomy, and their related degrees of similarity. Semantic similarity is evaluated using the information content approach, which shows a higher correlation with human judgment than the traditional similarity measures.