Using ontologies as queries in information retrieval

Current expectations from nowadays information retrieval systems (IRS) have grown beyond the “document contains these terms” requirement that was considered common sense 10–15 years ago. Nowadays systems are expected to return results that are relevant to the intended meaning of the query. In the general IRS usage scenario, the user is not really interested if the returned documents contain or not the words he entered in the query, but the user expects the returned documents to be relevant to the intended meaning of his query. The subject of analyzing the query (expressed as a sentence in natural language) in order to infer the meaning of the query, has spawned lots of research publications in the last 10 years, but without any large impact in the user search experience. We propose a system (the OntoSense User Interface) that guides the user in providing the query directly as a query ontology, thus providing the meaning of the query directly to the IRS. This approach is especially useful in mobile platform information retrieval scenarios where the query sentence is usually too short (very few words are used) for the IRS to match relevant documents.