Event-Concept Pair Series Extraction to Represent Medical Complications from Texts

Abstract

This research aims to determine an event-concept pair series as consequent events, particularly a Cause-Effect-concept pair (called ‘CEpair’) series on disease documents downloaded from hospital-web-boards. CEpair series are used for representing medical/disease complications which benefit for Solving system. Each causative/effect event concept is expressed by a verb phrase of an elementary discourse unit (EDU) which is a simple sentence. The research has three problems; how to determine each adjacent-EDU pair having the cause-effect relation, how to determine a CEpair series mingled with non-causeeffect-relation EDUs, and how to identify the complication of several extracted CEpair series from the documents. Therefore, we extract NWordCo-concept set having the causative/effect concepts from EDUs’ verb phrases including a support vector machine to solve each NWordCo size. We apply the Naïve Bayes classifier to learn and extract an NWordCoconcept pair set as a knowledge template having the cause-effect relation from the documents. We then propose using the knowledge template to extract several CEpair series. We also apply the intersection of the NWordCo-concept sets to identify the commoncause/effect for representing the complication-development parts of the extracted-CEpair series. The research results provide the high percent correctness of the CEpair-series determination from the documents.