INFORMATION EXTRACTION AND DOCUMENT CLASSIFICATION OF MEDLINE

Abstract

Biomedical literature indexing and classification is time-consuming process that is prone to inconsistencies. Pubmed and Medline repositories are growing at the rate of 500000 articles per year. Therefore, an automated system which could correctly determine the relevant keywords of the abstracts retrieved from Pubmed , is needed. In this paper we developed classifier, to classify Medline documents published in between 2000-2010.

We are restricted to Type 2 diabetes millets related literature only to decrease the size of corpora .Our objective was to investigate the benefits of using the MeSH controlled vocabulary as features to represent MEDLINE abstracts.. We developed and evaluated a classification model based on global relevant weighting schema to reduce the dimension Our algorithm out performed when compared to two popular Bayes and KNN classifiers and gives 20 % high accuracy than standard classifiers in worst cases.