Possibilistic Information Retrieval Model Based on a Multi-terminology.

We proposed in this paper a new approach for information retrieval intitled Conceptual Information Retrieval Model (CIRM). Our contribution is to exploit possibilistic networks (PN) and a multi-terminology in order to extract and disambiguate terms and then to retrieve documents. The two measures of possibility and necessity were used to select the relevant concept of an ambiguous term. Thus, the user query and unstructured documents are described throught a conceptual representation. Concepts were then filtered and ranked. Finally, a possibilistic network was exploited to match documents and queries. Two biomedical terminologies were exploited which are the MeSH thesaurus (Medical Subject Headings) and the SNOMED-CT ontology (Systematized Nomenclature of Medicine of Clinical Terms). The experimentations performed with CIRM on the OHSUMED corpus showed encouraging results: the improvement rates are +43.18% and +43.75% in terms of Main Average Precision and Normalized Discounted Cumulative Gain when compared to the baseline.