Bio-molecular event trigger extraction by word sense disambiguation based on supervised machine learning using wordnet-based data decomposition and feature selection

Abstract

Event extraction is a task of extracting detailed biological phenomenon from biomedical literature. This task needs to extract the words which represent the biological phenomena in terms of natural language in the form of text data. Trigger words are often ambiguous and can convey different meanings in different contexts in biomedical data. We use a supervised approach to disambiguate senses of such ambiguous trigger words. In this paper, we propose a supervised machine learning approach for Word Sense Disambiguation using Genetic Algorithm (GA) for feature selection. We take help of Wordnet dictionary to disambiguate words. Our experiments are applied on BioNLP-2011 datasets and we find recall, precision and F-score of 70.71%, 83.70% and 76.66%, respectively, in bio-molecular event trigger extraction.