Domain Independent Event Extraction System Using Text Meaning Representation Adopted for Semantic Web

Abstract

Given a text the proposed work is to identify an event of interest from the text and represent the result as metadata for the semantic web. As the existing event extraction systems are based on machine learning techniques and they do not assimilate the semantic analysis, there is a need for event extraction system that incorporates semantics. The proposed work provides a framework for identifying events from the text documents in the web by considering its text meaning representation (TMR) and adding the information as metadata. The system comprises four segments namely, identifying event triggers, event argument, and event properties and annotating the text with the event information obtained from first three segments. Event triggers are recognized using conditional probability. Event arguments are identified from TMR of the input text. Event properties are extracted from syntactic analysis of text and TMR. This information is represented in the OWL and RDF format and attached as metadata to the text in the corresponding web document, so that it can be accessed by queries that need information about the events. The proposed work mainly deals with the case of the single event represented in more than one sentence and adding event information to the semantic web.