Automatic Generation Of Sci-tech Innovation Enterprise Evaluation Report Based On The Coverage Of Knowledge Graph

The Science and Technology Innovation Enterprise is centered on technological innovation,with strong R?D capabilities and stable profitability.The evaluation of sci-tech innovation enterprises is conducive to market supervision to promote the healthy development of high-quality enterprises,better serve the national science and technology innovation strategy,and promote highquality economic development.At present,many institutions and analysts are researching sci-tech innovation enterprises and publishing sci-tech innovation enterprise research reports with professional content and reliable information.The evaluation text of the sci-tech innovation enterprise is to evaluate the sci-tech innovation enterprise in a certain period,which is time-effective.Financial practitioners need to write evaluation reports for sci-tech innovation enterprises regularly.The evaluation reports are written do not have a unified evaluation indicator system,the evaluation indicators are one-sided,and the format is not uniform.Therefore,it is of great significance and value to integrate the relevant information of sci-tech innovation enterprises from many sci-tech innovation enterprises and display them intuitively and automatically generate evaluation reports for sci-tech innovation enterprises.In recent years,with the in-depth research of natural language processing and the application of knowledge graphs,it is possible to automatically generate the evaluation report of sci-tech innovation enterprises.This article first studies the automated construction of the fusion knowledge graph of sci-tech innovation enterprises.This paper uses the Bi LSTM-Attention-CRF model to construct the knowledge graph of sci-tech innovation enterprises and the industry knowledge graph.Uses the entity link of sci-tech innovation enterprises and the embedding of the knowledge graph to represent the two kinds of knowledge graph fusion methods to generate the fusion knowledge graph of scitech innovation enterprises.Shows the information of sci-tech innovation enterprises and the upstream and downstream distribution of the industry chain.This paper then studies the generation of evaluation texts for sci-tech innovation enterprises based on the information coverage of the fusion knowledge graph of sci-tech innovation enterprises.This paper calculates the coverage rate of the fusion knowledge graph of sci-tech innovation enterprises in generating the evaluation text of sci-tech innovation enterprises.It proposes introducing the coverage rate into the Bi LSTMAttention model to generate the evaluation text set of sci-tech innovation enterprises.Make full use of the fusion knowledge graph information of sci-tech innovation enterprises in generating assessment texts for sci-tech innovation enterprises.At the end of this paper,we study the generation of the evaluation report of sci-tech innovation enterprises based on the evaluation template of scitech innovation enterprises.This article refers to the relevant national sci-tech innovation enterprise evaluation standards and builds a template for the evaluation report generation of sci-tech innovation enterprises.The semi-supervised text matching model is used to match the sci-tech innovation enterprise evaluation text set and the sci-tech innovation enterprise evaluation report template.According to the template,filter and sort the previously generated evaluation texts of sci-tech innovation enterprises.The sci-tech innovation enterprise evaluation report is finally generated.This article uses sci-tech innovation enterprise research reports,industry research reports issued by securities institutions and research institutes,and corporate annual reports issued by companies as data sources to automatically construct a sci-tech innovation enterprise integrated knowledge graph covering the industry information of sci-tech innovation enterprise and enterprise information on sci-tech innovation enterprise.This paper analyzes the distribution of the coverage rate p of the generated sci-tech innovation enterprise fusion knowledge graph for each target text.Using ROUGE and BLEU evaluation methods proves that the fluency and accuracy of the generated sci-tech innovation enterprise evaluation text are improved.Verify the effectiveness of the model.We constructed a template for the evaluation report generation of sci-tech innovation enterprises covering comprehensive indicators.Based on the sci-tech innovation enterprise evaluation template,the evaluation report of sci-tech innovation enterprises is generated with a unified format,comprehensive evaluation indicators,controllable content,smooth and accurate.