Research On Event Extraction Based On Sentence Framework And Parsing

Many cognitive scientists believe that memory is stored by the manner of event which is the unit for storage in the human mind. Simulating the manner of human memory and storing knowledge in the form of events in the computer has important significance for the realization of intelligent computer. Event extraction is an important research direction in the field of information extraction, and it is an important research direction of artificial intelligence too. Event as a representation of the knowledge has a profound impact on the field of the research question answering, natural language processing, information retrieval and automatic summarization.Events can be divided into meta event and topic event. Meta event is manifested as a sentence-level event in the text, namely the microscopic size event. Topic event is manifested as a chapter-level event in the text. The object of this thesis discussed is microscopic particle events, namely meta-event. For the definition of events have not form a unified standard currently, we use the definition of the event which is given by Lu-Chuan. The event is defined as a central affair element (i.e., event verbs of this article) and one or more peripheral affair element (i.e., peripheral role).Internationally, there are two main methods to research event extraction at present:based on pattern matching and based on machine learning. Although the method based on pattern matching can achieve good results in the field, its portability is far from ideal and the establishment of pattern requires a lot of manpower and resources. The method based on machine learning doesn’t need to build pattern. For event extraction, this method relies on the corpus and uses the thinking of text classification. The method has also made good results in event extraction.We proposes an improved event extraction algorithm which is based on parsing, and we proposes an event extraction algorithm which is based on sentence framework.The main contribution of this thesis are as follows:(1) Using the Word segmentation ICTCLAS2015 developed by Chinese Academy of Sciences and the parsing tool Parser developed Stanford University to preprocess the text. And then based on the characteristics of parsing which are summarized by predecessors and ourselves we propose an improved event extraction algorithm which is based on parsing. It can automatically extract the event and the event subject, object, time and other information from the text.(2) We revise and simply the sentence framework which is defined by Lu-Chuan, and establish a sentence framework table of event verbs. The sentence framework table is used to match sentence framework with the text which has been segmented, so as to extract the peripheral role of event.(3) For the shortage of the event extraction algorithm based on parsing, we proposes an event extraction algorithm which is based on sentence framework. The algorithm can extract the peripheral role of event well and make the peripheral role more precise.(4) We also proposes a matching sentence framework algorithm. The algorithm match sentence framework with the text which has been segmented and the text which has been parsed. If the match is successful, then using the sentence framework to extract peripheral role of event, else using the event extract algorithm based on parsing to extract event and its peripheral role.