Research On Knowledge Acquisition From Text Based On An Emergency Knowledge Model

Knowledge Acquisition from text has always been a rearch focus as a bottleneck in the field of Artificial Intelligence and Knowledge Base System development. Plenty of knowledge distribute everywhere in the text, also the same situation in emergency domain, there are plenty of knowledge such as professional knowledge of disasters, emergency responses and dynamic laws of disasters. However, Most of the existing methods are depend on Knowledge Engineer (KE), or as a way of assisting KE acquisite knowledge from text, as a result, most of them have a low efficience and narrow applicating field, also most of the knowledge they acquisite are simply concepts and relationships, and not applicable to our project.To deal with the above problems, this paper analysis the knowledge needs of the project and characteristics of texts in emergency domain, and finally presents a knowledge model of emergency to manage and represent the abstracts of disasters, and also a dictionary of logical relationships. Based on this model and dictionary, a method of extracting facts and relationship between them from texts of emergency field is proposed, combined pattern-matching method, part of Natural Language Processing technology and statistical methods. And also propose a method of representing the extracted facts and relationships in a way of graph, as a basis of extraction of knowledge pattern which is the subsequence task, in order to refine the existing knowledge, and finally support emergency decision-making.In this paper, a concept "General Fact" is proposed for the first time and a method of extracting General Fact from text, and a foundation is constructed for the extraction of rules and knowledge pattern. This paper also proposes using a dictionary of logical relationships to extract relationships between facts. The extracted knowledge can be applied to the Knowledge Base of the project directly and has some practical and theoretical value.