Research On Key Technologies Of Event Knowledge Graph For Dealing With Emergencies

Emergencies refer to natural disasters,accidents,public health incidents,and social security incidents that occur suddenly,endanger the society,and urgently need to be dealt with and responded to.Analyzing and predicting emergencies and taking corresponding measures to prevent and respond in time can effectively control the loss of personnel and property and the deterioration of the situation.Therefore,this article is oriented to the field of dealing with emergencies,establishing a complete information extraction system for emergencies,constructing an EKG in the field,and assisting domain experts in judgment and processing.The main work of this paper is as follows:(1)Research on event extraction methods based on multi-task learning.This article divides the event extraction work into two subtasks,event detection and event element identification.For event detection,we study the pre-training-based multi-task learning event detection method,adopt the word sense disambiguation task,and initialize the vector by fine-tuning the pretraining model to solve the problem of sparse corpus and polysemy of trigger words.For event element recognition,a joint extraction model of named entities and event elements is proposed,multi-task learning is performed through shared parameters,and vocabulary information is adopted to enhance the semantics of the text,clarify the word boundaries of event elements,and finally pass experiments.(2)Research on causality extraction methods for emergencies based on knowledge enhancement.This paper uses event causality as a starting point to study the evolution of events,and proposes a method of extracting event causality based on knowledge enhancement.This method judges the causality of the events in the text based on the completion of the event detection task.Through pattern matching,causal event sentences are extracted from a large amount of corpus to construct a causal association graph of word granularity,and input them as causal prior knowledge into the subsequent model to complete the causal relationship identification.(3)Establish an application verification system for EKG in the field of dealing with emergencies.On the basis of the above research content,combined with the event extraction model and the event causality recognition model to design and implement an event knowledge graph application verification system.The system mainly includes four modules: data collection and processing,data labeling,emergent event graph construction and emergent event graph application.Through event standardization and event fusion,the problems of inconsistency of time entities and repetition of similar events in the map construction process were solved.Finally,system testing and system page display were carried out.