The Similarity Calculation Of Emergency Based On Scenarios Frame And Dependency Tree

With the development of human civilization and social progress, material and spiritual civilization of human have get great development, at the same time, a variety of unexpected natural or public events occur frequently brought great loss to the community, how to take measures to deal with the incident effectively and timely is the focus of the government. Learn and analyze from the historical case of emergencies can help decision makers to get the historical experience and make the correct decision. Because historical cases of emergency exist mostly in text, how to get the similar historical case with the current disaster from the case base of the case base similar to the current disaster is in a hurry to be solved. Therefore it is needed to research the similarity calculation of emergency text.Text Similarity calculation has been mature, lots of domestic and foreign scholars has researched the problem. Such as, in foreign, Gerard Slaton and McGill proposed VSM, the vector space model, it use vector to express text, vector has lots of key words of the text, every key words has its weight, at last we can use the similarity of the vector to respect the similarity of the text; Hirschberg proposed a method to calculate the similarity of text which is based on LCS. In china, Jin Yaohong proposed a method to calculate the similarity of text which is based on contextual framework, it divided the text into three parts, field, scenarios and background; Jing Bo proposed a method to calculate the similarity of text which is based on the semantic understanding. Scholars have lots of research about the text similarity calculation, but most of them is based on sequence of the words, the words frequency or semantic structure of the text for ordinary text.Compared with ordinary text, emergency events text focuses on more the state properties, the causes, impact, decision-making of the event than the syntax and semantics of the theme of the text. In this paper, we take into account the characteristics of emergency events and proposed a new way to calculate the similarity of emergency events text. It is based on the scenarios framework and dependency tree-based patterns, the main thought of the method is: to measure the similarity of the text by using the similarity of the scenarios framework. The emergency scenario framework model, ESF, divided the emergency event into four parts," event, bearer, activity, derivative disasters ESF ", and then we extract information from text through dependency tree-based patterns and calculate the similarity through the scenarios frame.At last, compared with SVM method and LCS method, we use experiment to prove the effectiveness of the method.