Biological text mining for extraction of proteins and their interactions

Abstract

Text mining techniques have been proposed for extracting protein names and their interactions. First, we have made improvements on existing methods for handling single word protein names consisting of characters, special symbols, and numbers. Second, compound word protein names are extracted using conditional probabilities of the occurrences of neighboring words. Third, interactions are extracted based on Bayes theorem over discriminating verbs that represent the interactions of proteins. Experimental results demonstrate the feasibility of our approach with improved performance. © Springer-Verlag Berlin Heidelberg 2005.