Event recognition from information-linkage based using phrase tree traversal

Abstract

In this paper we present an approach to extracting significant events from digital documents. OpenNLP syntactical parser for English is used for generating parse trees from the sentences, followed by the extraction of events from the parse trees using tree traversal algorithms. The extraction system is developed and tested on 50 sentences from terrorism documents of The Federation of American Scientists (FAS). The results showed that with this technique we can achieve high recall and precision yielding accuracy of 89.68 recall and 78.44 precision with an overall performance of 83.66 in term of F-measure.