Research On Key Technology Of Event Extraction Based On Frame

With the popularity of computers and the rapid development of Internet, more and more information flooded in the network. How to extract information accurately and timely from a large amount of disordered data, sort and organize into the form easy to retrieve, is a pressing problem to solve, information extraction is based on this goal. Event extraction is an important part of information extraction; it extracts events which users have interest in from the unstructured document, and describes as a structured form for querying and analyzing further by users.The paper uses natural disasters events as subjects, studies the information feature in this field, extracts and merges the information of natural disasters events. Firstly, the paper filters the low-frequency words and high- frequency words in documents by pre-processing the experiment corpus, and then collects the feature words of events flank by using document clustering. After a certain amount of manual processing, the event frames of natural disaster have been defined, which will guide the event extraction. In the specific steps of event extraction, we define extraction template, that is, a set of extraction rules of relevant events firstly, then use these rules to do template matching and information extraction on the test corpus. We extract the natural disasters event's flank information such as time, location and result of casualties, and then merge the information belonged to an event according to the principle of time-space consistency.The experimental results show that the method of event extraction based on frame in this paper is feasible. The result of extraction on natural disaster events is satisfactory.