Research On Text Selection System Of Thermal Power Field Based On Knowledge Graph

With the development of computer and Internet,the society has entered the era of intelligent information,and the information of various industries is growing exponentially.In recent years,the combination of new technology and professional field has become a research trend,which provides people with a lot of convenience.With the rapid development of knowledge graph,data information can be associated to improve the utilization of data,and knowledge mining can be better carried out.Most of the existing knowledge graphs are oriented to the encyclopaedia knowledge domain,and there are few knowledge graphs for specific domains.Therefore,this paper studies the construction of knowledge graphs and text retrieval ranking for specific domains.The main research work is as follows:(1)Domain knowledge graph construction.A Bi LSTM-CRF named entity recognition model based on Bert is proposed.Before building the knowledge graph,the task of named entity recognition is studied.It is proposed to add the domain specific knowledge data to the Bert pre-training model,and then connect Bi LSTM-CRF to complete the task of named entity recognition.The model adds the Bert pre-training model to the traditional Bi LSTM-CRF model,which improves the model’s understanding of the context and can better recognize the entity.Through experiments on people’s daily data set and specific domain data set,the accuracy of named entity recognition is higher than other comparison algorithms,and the experiment proves the effectiveness of the method.(2)Improve the text selection model of NSGA-II.In this paper,according to the domain knowledge graph,knowledge reasoning and knowledge representation are carried out for the input sentences according to the professional knowledge,and a certain amount of professional description can be generated.In order to make the generated text more targeted,the multi-objective optimization algorithm is applied to the text sequence sorting.In this paper,the NSGA-II optimization algorithm is improved,and the differential mutation strategy is proposed to replace the NSGA-II mutation strategy,and the control factor in the differential mutation strategy is adjusted adaptively,so that the convergence and distribution of the algorithm are effectively improved.Experiments show that the improved algorithm is superior to similar algorithms in GD,IGD and SP.(3)Based on the knowledge map of specific fields,this paper constructs a text selection system platform in the field of thermal power.Through the analysis of project requirements and system characteristics,the main function of the platform is designed.Combined with knowledge graph visualization and text selection,the search function,text display function and mapping visualization function are designed.The system architecture is designed,which is divided into data layer,knowledge graph layer and application layer.Through the system running test,the basic design function is realized.