A Willing Events Identification Method

Events extraction is an important part of information extraction, at present, most events extraction aim at extracting events themselves but do not consider whether the events have already happened. This paper focuses on the events named willing events which have aleady proposed and made sure to happen but not yet happened currently. There exist quite a number of willing events in many areas, and those can provide a lot of information, thus, carries on the extraction is very meaningful. This paper presents a new feature selection method based on Dependency Parsing. Dependency parsing is used to find the syntactic relations among the words expresses willingness, event denoters and other words. Then these features are used in machine learning to classify the events, and finally achieve the identification of willing events. The experiment shows that, aiming at willing events, the features found based on dependency parsing do better than those features used in traditional events identification.