The Design And Implementation Of Knowledge Graph Based On Information Extraction For Airport Emergency Rescue Plan

Airport emergency rescue plan is the preformulated disposal plan for emergencies in the area of airport,which is the basis for airport to deal with emergencies.However,most airport emergency rescue plan are presented in the form of text,which is not intuitive and inconvenient for users to look up.Knowledge Graph is an efficient knowledge base,which can show knowledge intuitively and access knowledge quickly.The design and implementation of knowledge graph based on information extraction for airport emergency rescue plan aim to extract the key entity and relation from airport emergency rescue plan and take advantage of the airport emergency rescue plan efficiently.It is of great significance to improve the safety level of the aviation area.This paper focuses on methods of constructing knowledge graph for airport emergency rescue plan,including relation extraction and knowledge graph completion.The following work is carried out from two aspects:1.Aiming at relation extraction,the method of relation extraction fail to consider the correlation between head entity and words in the process of dealing with the overlapping triple.This paper proposes a joint extraction of entity and relation with entity enhancement to solve the problem.The method extracts triple under the different relation.Firstly,the head entity of triple is recognized.Then,the entity enhancement mechanism is adopted to get the correlation between the head entity and words to improve word embedding.Finally,the tail entity is recognized under different relation,solving the overlapping problem of triple.The experimental results on the airport emergency rescue plan show that the joint extraction of entity and relation with entity enhancement method can effectively learn the correlation between the head entity and words,improving the extraction of triple significantly.2.Aiming at knowledge graph completion,existing graph completion methods can not learn the structure of knowledge graph when when the knowledge graph is small.This paper proposes knowledge graph completion method based on pretrained language model to learn the text information contained in the graph.Splicing a triple into a sentence and complete link prediction by transforming task into the classification of judging whether the sentence is a positive sample or a negative sample.In addition,the entity token is added into the triple sentence to learn the entity information of the sentence.The multi-task learning mechanism is adopted to construct the sentence-level features for link prediction task,entity-level features for relation prediction task,and triple-evel features for relevance ranking task,which enhances the link prediction task.The experimental results show that the proposed method outperforms other methods of knowledge graph completion.3.Based on the above works,We design and implement a knowledge graph system for airport emergency rescue plan,including functions of storing graph,looking up the graph,and displaying the graph.It is used to look up the airport emergency rescue plan for users efficiently.