Research On The Extraction Of Dispose Knowledge From Text In Emergency Domain

Disposal Knowledge in Emergency Domain describes the relation among emergency situation and its disposal methods, which helps the workers to understand the current situations of the emergency events and predict its trends, and also provides suggestions to decision makers, thus is one of the most important kinds of knowledge within the knowledge base. Texts contain lots of professional information, rules and measures relevant to emergencies, the wide usage and rich in information make texts a desired knowledge source. This thesis is the research on how to extract Disposal Knowledge in Emergency Domain from texts.Judging from the current research, several problems exist when extracting knowledge from texts. First of all, the most popular method of the text information extraction is still the manual extraction, which is inefficient and unable to process mass texts, thus causes insufficiency finally. Although researches on auto-extracting methods already appeared, the process is always lack of domain model constraint, the content give priority to extract the fact, relationship is seldom covered. In addition to that, difficulties in understanding Chinese natural language make the modeling and extracting an even hard job, causing the inefficiency of knowledge acquisition.In view of the above questions, this thesis researches on how to automatically extract disposal knowledge from huge texts in emergency domain. Dictionary driven and pattern match are the basic method to be used.Analyses on the common expression of causal relation and knowledge demand in the field of emergency are present firstly, based on which a regular express of knowledge is expressed. Main work of this thesis is to extract disposal knowledge from Chinese natural language texts. Within the process, the domain dictionary including Domain Knowledge Dictionary and Task Knowledge Directory are constructed and follows the design of rules for extracting sentences containing the knowledge we need from texts. Based on the analysis of the structure of the sentences, causalities are extracted; domain knowledge is acquired from those causalities and the disposal knowledge is composed. An experiment is also introduced to evaluate the method.According to the result of the experiment, the method proposed in this thesis is able to extract most of the target knowledge that should be recognized manually, the recall rate and precise rate of the method are acceptable. The research of the method not only enriches the knowledge extraction related research, but also expands the research of text information extraction to the emergency management field, which also helps to improve the degree of emergency decision-making process.