Case-based Reasoning Research For Emergency

In recent years,the frequent occurrence of various emergencies all over the world has brought influence and harm to society harmony and people happiness.Combine the theory and technology of information technology,emergency management and relative disciplines to mine the information of emergencies.And research on case-based reasoning of emergency which has important theoretical and practical significance for text analysis,knowledge discovery,emergency management,public opinion monitoring and so on.Therefore,this thesis focuses on the process of case-based reasoning,and carries out the following research work:(1)Research on emergency case representation and information extraction.Firstly,the relationship between case representation and case-based reasoning is analyzed.Combining the characteristics of emergencies,a case representation method based on frame and production is proposed,which lays a solid semantic foundation for case-based reasoning of emergency.By analyzing the data sources of emergencies,the linguistic style concept is integrated into the rules of information extraction.Combining the terminology with rich domain information,this thesis proposes an information extraction method based on style and terminology.The structuring of data is completed and the case library of emergencies is constructed.Experiments carried on structuring data and CEC set have proved the validity and feasibility of the information extraction method.(2)Research on case-based reasoning method based on semantic similarity of words.Firstly,the relationship between case retrieval and case-based reasoning is analyzed,and the retrieval scheme of emergency attributes is determined.Taking the text attributes of emergency as the research object,the semantic similarity calculation methods of words are studied based on corpus and knowledge base respectively.Combine context information such as character,words and Ngram text sequences to enhance word embedding.On this basis,SSW_CCKM(Semantic Similarity of Words Computing by Combining Corpus-based and Knowledge-based Methods)method is proposed based on HowNet and Tongyici Cilin word semantic similarity algorithm;The validity of this method is verified on PKU 500 public dataset.(3)Research on case base maintenance method of emergency.Firstly,the relationship between case base maintenance and case-based reasoning is analyzed,and the case base maintenance scheme of emergency is determined.Introduce the idea of social network degree centrality into Kmeans text clustering and propose DC_Kmeans text clustering to identify noise cases and special cases.The concept of utility value in decision analysis is introduced to determine the value of each case,and the case base maintenance method of KUCBM is proposed,so realized the case base maintenance of source cases.Through the final data simulation,the effectiveness and feasibility of the KUCBM method proposed in this thesis are verified.