Extraction of Problem Events from Web Documents to Construct Cause-Effect Loop

Abstract

This research aims to extract problem events, particularly cause-effect concept pair series with explanations by several simple sentences with causative/effect concepts, from web documents of drug addiction. The extracted problem events are used to construct cause-effect loop which benefits for the problem analysis in the solving system. The research has three problems; how to determine the cause/effect event concepts expressed by verb phrases having a problem of the overlap between causative-verb concepts and effect-verb concepts, how to determine the series of cause-effect concept pairs with the causative/effect concept boundary consideration, and how to determine the feedback-loop of cause-effect concept pair series. Therefore, we apply the event rate to solve the overlap problem. We then propose using N-WordCo to determine the cause-effect concept pair series and also use a cue-word set to solve the feedback-loop. The research results provide the high precision of the problem event extraction from the documents.