AUTOMATED KEYWORD EXTRACTION FROM BIO-MEDICAL LITERATURE WITH CONCENTRATION ON ANTIBIOTIC RESISTANCE

Abstract

The explosive growth of bio-medical literature makes it increasingly difficult and time consuming to keep up with newly discovered and published information. The extraction of knowledge from papers is critical in enabling computational analysis of biological data. In the last decade, tremendous effort has been put into development of automated and semi-automated tools for knowledge discovery and extraction from text, as an alternative to monotonous and time-consuming manual processing. This thesis research was focused on determining whether minor human supervision can improve the process of automated bio-medical text annotation. One of the main outcomes of this study is a tool that requires minimal effort and time from scientists to reach high precision in semi-automated annotation. The task we targeted is the extraction of keywords related to antibiotic resistance in bacteria. The tool is based on a machine learning algorithm that is retrained several times to achieve the best accuracy.