Design And Implementation Of Event Detection System With Integrating Dependency Information

With the development of the field of natural language processing,event detection has become an important information extraction task in natural language processing.The result of event detection is an important knowledge required for other tasks.And it is an important knowledge supplement for applications such as text search and event knowledge graphs.In recent years,methods based on dependency syntax information and graph convolutional neural networks(GCN)have been widely used in event detection tasks.For event detection,graph convolutional networks based on dependency arcs can capture the dependency grammatical information between candidate trigger words and arguments.However,the existing GCN methods based on syntactic graphs have problems of graph imbalance and information redundancy.In order to capture the important information in the graph and further refine the information,this paper proposes a multi-graph convolutional network with jump connections,referred to as MGJ-ED.In the multi-graph convolutional network module,a core subgraph separated from the dependency graph.The dependency graph uses GCN to select important syntactic information of one-hop neighbor nodes in breadth.In the jump connection architecture,the expressions of different attention scores of different GCN layers are aggregated,and the attention weight of the grammatical information of neighbor nodes with different hops is learned in depth.Compared with other event detection models,the model in this paper improves the recall rate by 2.5%and the F1-measure score by 2.0%on the widely used ACE 2005 data set.The experimental results show the superiority of the method in this paper.The papers written for the above innovations have been accepted and published by international conferences.In this paper,an event detection system that integrates dependent information is designed and developed.The system uses a front-end and back-end separation framework design and uses the MongoDB database system to realize cutting-edge knowledge browsing and text event detection functions.At the same time,this paper applies the studied algorithm model to the event detection system.The relevant tests on the event detection system that integrates dependent information have all passed,and the system response time is within 500ms,and the event detection result can be given within 1s.The event detection system that integrates dependency information can be used in practice.