IntEx: A Syntactic Role Driven Protein-Protein Interaction Extractorfor Bio-Medical Text

摘要

In this paper, we present a fully automatedextraction system, named IntEx, toidentify gene and protein interactions inbiomedical text. Our approach is based onfirst splitting complex sentences into simpleclausal structures made up of syntacticroles. Then, tagging biological entitieswith the help of biomedical and linguisticontologies. Finally, extracting completeinteractions by analyzing the matchingcontents of syntactic roles and their linguisticallysignificant combinations. Ourextraction system handles complex sentencesand extracts multiple and nested interactionsspecified in a sentence.Experimental evaluations with two otherstate of the art extraction systems indicatethat the IntEx system achieves better performancewithout the labor intensive patternengineering requirement.