| Tobacco leaf grading work is directly related to cigarette manufacturing and farmers’ income,so it is particularly important to improve the efficiency and accuracy of tobacco leaf grading work.However,under the existing conditions,tobacco grading knowledge mainly exists in the form of books and manuals,making it difficult for relevant staff to quickly obtain accurate knowledge when encountering problems in actual operation or training.To solve this problem,knowledge graph technology is used to organize and store knowledge in the field of tobacco leaf grading,enabling relevant personnel to more efficiently acquire and apply this knowledge.At the same time,deep learning models are used to process natural language queries to understand the semantic information of the problem.Based on the above technologies,a tobacco leaf grading intelligent question and answer system has been designed and implemented to provide knowledge answers for staff.This Q&A system has important practical application value,not only helping to promote the intelligent and digital development of the tobacco industry,but also providing an innovative solution for the tobacco industry to deal with the limitations of existing knowledge acquisition channels.The main research contents and results are as follows:(1)Research on the construction of tobacco leaf grading knowledge graph.The tobacco leaf grading knowledge graph pattern design is completed through a top-down approach.In response to the high cost,time-consuming,and labor-intensive problems of traditional knowledge graph construction methods,a method based on Bert+Text CNN+Res Net+Caps Net+BI-LSTM+CRF is proposed for constructing the data layer of the tobacco leaf grading knowledge graph,accomplishing the entity recognition task of tobacco leaf grading.The entity recognition value of this method is96.9%,significantly improving the performance compared to other related models.Triple extraction is implemented using the annotation method for extracted entities.Entity alignment is used to solve the ambiguity problem between entity concepts based on the dataset characteristics.Finally,the Neo4 j graph database is selected to store the knowledge and display the tobacco leaf grading knowledge graph.(2)Design of a tobacco leaf grading intelligent question-answering method based on knowledge graph.Building upon the constructed tobacco leaf grading knowledge graph,a joint extraction approach for question intent classification and entity recognition is proposed.The identified entities and intentions are integrated to form the triple of the problem,and the Cypher language is used to retrieve the local Neo4 j database.Experiments show that the average value of the question entity recognition method reaches 97.7%,and the question intention classification is 97.6%.(3)Design and implementation of an intelligent question-answering system based on the tobacco leaf grading knowledge graph.Utilizing the tobacco leaf grading knowledge graph as the data foundation,the Flask application framework and MVC response pattern are integrated to construct the front and back end of the tobacco leaf grading intelligent question-answering system,enhancing the tobacco leaf grading knowledge service system and advancing the digital development of the tobacco industry. |
