Event Detection Driven Approach for Extracting Information from Internet Documents

Abstract

In this paper, we propose event detection driven intelligent information extraction by using neural network paradigm. A back-propagation (BP) approach learning algorithm is adopted in the proposed system to train the event detector. In order to detect the potential events in documents effectively, we apply an part-of-speech tagger to help select nouns as feature words. Furthermore, unrelated nouns are filtered by the analysis based on document frequency distribution. At last, selected nouns are conceptualized into concepts using Ontology. These concepts are thought to characterize documents mostly. In the experimental results, we got high accuracy both in the inside testing and outside testing of Internet documents. By means of the well-trained event detector, the information extraction task can be applied in wider domains by the aid of intelligent event detection. This event detection technology is introduced in this paper for the application of E-mail delivery.