Improving Semantic Information Retrieval Using Multinomial Naive Bayes Classifier and Bayesian Networks

This research proposes a new approach to improve information retrieval systems based on multinomial naive Bayes classifier (MNBC), Bayesian networks (BNs), and multi-terminology which includes MeSH thesaurus (Medical Subject Headings) SNOMED CT (Systematized Nomenclature of Medicine Clinical Terms). Our approach, is entitled improving semantic (IMSIR), extracts disambiguates concepts retrieves documents. Relevant ambiguous terms were selected using probability measures biomedical terminologies. Concepts are also extracted an MNBC. The UMLS (Unified Medical Language System) was then used filter rank concepts. Finally, we exploited network match documents queries conceptual representation. main contribution in this paper combine supervised method (MNBC) unsupervised (BN) extract from queries. We propose filtering the order keep relevant ones. Experiments IMSIR two corpora, OHSUMED corpus Trial (CT) corpus, interesting because their results outperformed those baseline: P@50 improvement rate +36.5% over baseline when used.