Automatic Category Recognition of MeSH Terms through Comparison of their Occurrence Frequency in Relevant and non-Relevant Documents

Due to dynamic of terms, their classification is challenging. The current research aims at determining the usability of a model for automatic recognition of MeSH terms categories through measuring their occurrence frequency within relevant and non-relevant document corpuses from PubMed. This is a descriptive research that uses the document analysis method. MeSH and PubMed were used to collect research data. The significancy of these resources confirms their validity. 18164 MeSH-term and 163226 PubMed documents were selected. The both of these amounts are greater than what Cocran function suggests. Eleven document corpuses were retrieved from PubMed. The relative ocurrence frequencies of MeSH terms within each corpus were determined. The results were compared with the real category of MeSH. In additions, the categories of 1 percent of MeSH terms were determined by experts in medical domains. The frequency distribution method was used for statistical description of data. Data were also analyzed through T and Chi-Squar tests in SPSS. Each document of PubMed on average belongs to three MeSH categoris and most of Mesh terms occurred in all corpuses. The results confirm that the suggested method increases the probability of MeSH category recognition. The performance of the method depends on the subject category of MeSH Term and ranges between 3 to 67 percent. The findings also show that the medical expertises determination on the subject category of MeSH Terms is compatible with the real categories of MeSH tree. The compatibility of the subjective and objective methods for the subject category recognition depends on the knowledge area. The subjective categorization is a quite cognitive task and roots in human environmental experiences. This is why the machine depended models are not able to simulate that process.