An event extraction method by semantic role analysis

Abstract

The paper does study on event extraction from news on the internet by the method of semantic role analysis. We annotate the sentence in the news from argument annotator, extract the argument structure of the head verb, convert the arguments to specific semantic roles of the verb, and then match the semantic roles to the event elements. This paper puts forward and studies on how to use VerbNet and SemLink resources to match the verb's arguments and event elements. The experiment was carried out on the 1000 news corpus crawled from the web, and the result shows that the F value is up to 70.6% and has certain application value.