Research And Implementation Of Question Answering Technology For High-speed Railway Operation Based On Knowledge Graph

With the rapid development of high-speed railways,the component structure and operation mode of EMUs have become more and more complex,and the requirements for emergency handling capabilities of EMUs have become more and more strict.However,practitioners in this field often need to check repeatedly in their daily operation and maintenance work.There are a large number of historical instruction manuals,and the entire processing process is even highly dependent on the operator’s previous experience,resulting in low work efficiency,and it is difficult to meet the requirements for rapid failure analysis and processing of EMUs.Although the traditional method of text retrieval by keywords can locate paragraphs,However,the retrieval results are fragmented and lack organization,often resulting in incomplete retrieval,negligence and omission,which also exacerbates the reduction of operation and maintenance work efficiency.With the development of knowledge graph technology,more and more fields such as medical care and agriculture have begun to use knowledge graphs to store and organize information resources,and use this as data support to realize efficient consulting and question answering,intelligent search applications,and high-speed railway operation.It is also possible to learn from the above technologies,use the knowledge graph to organize data in the field,and provide accurate,convenient and efficient consulting and question answering services,thereby improving the work efficiency of practitioners.The main work of this paper is as follows:(1)The construction method of knowledge graph in the field of high-speed railway operation is proposed: this paper constructs a knowledge graph model layer suitable for the field of high-speed railway operation by effectively analyzing the knowledge data related to high-speed railway operation.The corresponding data acquisition strategy is developed.After data cleaning,data such as fault handling,work items,and EMU equipment are formatted according to the definition of the mode layer.Finally,the CREATE statement and LOAD CSV statement are used for data in different formats.Two methods are used for data storage,and the construction of a knowledge graph in the field of high-speed railway operation has been preliminarily completed,including a total of 1,241 entities and 2,372 entity relationships.(2)Aiming at the lack of labeled corpus in the field of high-speed railway operation,a method for generating questions based on knowledge base data is proposed.This method constructs multiple question templates for each type of intent after completing the induction of domain intent.After random selection of templates and entity filling,a high-quality domain question dataset can be generated for model training.(3)For the problem that the BERT model considers insufficient semantic information in the intent recognition task,it is combined with the Text CNN model to mine more semantic features,and a character embedding-based BTT(BERT-Text CNN-Token)and sentence embedding-based BTS(BERT-Text CNN-Sentence)are two intent recognition models.Compared with the BERT model,BTT considers local text features,BTS considers high and low level semantic information,and has a significant impact on the model’s main features.Several sets of experiments are carried out on the main parameters of the model,such as the size of the convolution kernel and the number of coding layers.The experimental results show that the performance of the two models proposed in this paper is improved compared with the BERT model,and it is more sensitive to keyword information.For the task of entity recognition,in order to build a more lightweight question answering system,a method of using the Han LP tool library and adding a professional dictionary is proposed to achieve efficient and accurate entity recognition without annotation.(4)Integrated and implemented the Question answering system for high-speed railway operation: Based on the research results of the above work,a high-speed railway operation consultation Question answering system was implemented using the Flask and Vue frameworks for front-end and back-end development,which can provide users with possible answers within the scope of the knowledge base.An interpretive Question answering service.