EVENT GRAPH CONSTRUCTION METHOD ON NATURAL DISASTER RESEARCH

Abstract. The analysis of natural hazards is still lacking in real-time and comprehensiveness. To enhance the intelligent analysis of disasters, reduce natural disasters, and provide more timely and accurate disaster warnings. The method of constructing a natural disaster research event graph using the abstracts of research literature related to natural disaster analysis as the data source is proposed. First, a rule-based matching algorithm combined with syntactic analysis is used to extract the events of the disaster analysis process and spatio-temporal information from the abstract; Then build an event storage model based on Neo4j to store the event chain and related information; Then use Word2vec to convert events into word vectors, and define the event similarity as a linear sum of the object similarity and predicate similarity of the two events, combine event similarity setting equation to calculate core nodes and perform event generalization and fusion of graphs. Based on the above method, soil erosion is used as an example to construct an event graph and provide the basis for decision making services for intelligent analysis of natural disasters.