Design And Implementation Of A Question And Answer System For Rice Diseases And Pests Based On Knowledge Graph

Rice is one of the most important food crop in China.In China,rice’s planting area and yield are among the top in the world,carrying the heavy burden of China’s historical inheritance and the happiness of the people.However,nowadays,the problem of diseases and pests has seriously affected the quality and yield of rice.Planters are not only unable to quickly obtain planting assistance from books and literature,but also unable to accurately obtain planting knowledge through internet search engines.In response to the above issues,this thesis will study,design,and implement a rice pest and disease question answering system,providing accurate and fast answers for rice planting personnel.The main work of this thesis is as follows.(1)Research on the Construction of a Knowledge Graph of Rice Diseases and Pests.Aimed at the complex relationship of knowledge in the field of rice pests and diseases,the conceptual level design of rice pests and diseases was carried out by combining top-down and bottom-up methods;In response to the difficulty in constructing a knowledge graph caused by the heterogeneity of rice pest and disease knowledge from multiple sources,unstructured knowledge is extracted using the BIOES sequence annotation method and the BERT-Bi LSTM-CRF model for entity extraction.This study is based on the BERT-Bi LSTM-CRF named entity recognition model.In response to the problem of low robustness and accuracy of the model,a network improvement strategy incorporating adversarial training and introducing noise is proposed,which utilizes shared information to optimize entity extraction tasks.In addition,a hierarchical learning rate strategy and an automatic learning rate decay strategy are introduced to improve the model’s entity recognition performance.The results showed that compared with the traditional BERT Bi LSTM CRF,the F1 value of this model increased by 2.34%,reaching 91.18%.(2)Research on rice disease and pest knowledge Q&A based on knowledge graph.To address the issue of insufficient and inaccurate Q&A on traditional search engines,a knowledge graph based Q&A on rice pests and diseases is proposed;In response to the problem of short text and few semantic features in question sentences for rice pest and disease users,the BERT Text CNN model is used to extract text feature information and achieve understanding of question intentions;To address the problem of problem entity recognition,a BERT-Bi LSTM-CRF model integrating adversarial training is proposed to recognize question sentence entities.For the answer query question,the Cypher statement is used to query the knowledge map of rice pests and diseases.The results showed that the F1 value of question intention understanding reached 97.81%,and the F1 value of answer query reached 79.32%.(3)Design and Implementation of a Knowledge Graph Based Question and Answer System for Rice Diseases and Pests.Developing and building a rice pest and disease question and answer system through the Django framework meets the needs of planting personnel to efficiently and quickly acquire rice planting knowledge,providing experience for the prevention and control of pests and diseases in information agriculture.