Two Approaches for Biomedical Text Classification

Automatic text classification systems can be especially valuable to biomedical researchers who seek to discover knowledge from terabyte-scale biomedical literatures. Different from the general domain, biomedical literatures contain a large number of named entities, complicated session structures and rich ontology resources. Taking these features into account, two approaches for biomedical text classification are presented, i.e., concept expansion and Meta-classification. Concept expansion is a method that introduces concept features using biomedical named entity recognition. Meta-classification is to combine the classification results of different parts of the full-text article and ontology resources using a Logistic regression model. The experiment results on the test set of TREC 2005 genomics track categorization task show that these techniques can improve the performance of the classification system consistently for all the classes.