Research On Event Extraction Technology In The Field Of Unexpected Events

With the explosion of network information, the number of information is growing exponentially. What followed was a serious problem that how to accurately and effectively extract event information of interest to the user from the information in these large and no structure. In order to extract the different levels and granularity of information from different sources of information, people study a variety of information extraction technology. Event is usually considered to be happening at a particular time in a particular place. Identification and extraction of C hinese text event is designed to use auto matic extraction technology to extract event information from a variety of unstructured text datas, its purpose is to develop a series of event-based information organization technology.With the development of information technology, corpus becomes an important tool for natural language processing and research, which is a collection of computer-readable texts, contains a representative sample that people interested. We built event-based Chinese corpus CEC-2. In the paper, the event information extraction is based on this corpus.The purpose of the event extraction is to extract the event that the user interests from unstructured documents, while descript in structural or semi-structured form, for users analyze and use further. Event identification is also often used as a sub event extraction task. Event identification is the basis for event extraction, which directly influences the outcome of the event extraction. In the paper, event recognition is based on the event trigger word. In the event ontology, a sentence that describe an event must contain a trigger word, a sentence that contains a trigger word, it ’s not necessarily that the sentence describe an event. In this paper, we obtain event recognition rules based on the frequent subtree of dependency syntax tree.Event elements extraction is another core event extraction task. In the paper, the task is to identify the elements of true events from the identified event. Event consists of six elements, our task is to identify the rest of the five elements e xcept the element of Language performance. We first mining frequent subtrees from a large number of corpora, and then get mode, which will make event elements be extracted. Experimental results got better recall R, precision P and F measure.