Automatic Question Answering System For Grape Diseases And Pests Based On Knowledge Graph

China has attached a lot of attention in agricultural informatization.Grapes have a large area of plant life and a high yield in China.The appearance of grape diseases and pests has a great influence on the yield and quality.The knowledge of grape diseases and pests exists in the form of books and documents,so grape growers cannot get them quickly and accurately.In this study,the knowledge graph is used to organize the domain knowledge of grape diseases and insect pests,and the deep learning model is used to understand the semantic information of natural language questions.The knowledge graph-based system for automatically answering the questions of grape and insect diseases has been designed and implemented.This study is meaningful for the intelligent development of the grape industry,because the knowledge service of grape industry in China is not yet perfect at present.The main study contents and achievements are as follows.(1)Research on the method of constructing knowledge graph of grape diseases and insect pests.According to the thinking of top-down and bottom-up,the conceptual model of grape diseases and pests’ knowledge graph was constructed,aiming at the lack of open source knowledge graph.That is,from the top down to comb the grape diseases and pests’ domain model,from the bottom up to supplement the notion of model.A method to build the data level of knowledge graph of grape diseases and insect pests based on BiLSTM-CRF was proposed to solve the problem of high cost of manual construction of knowledge graph.The text information is extracted automatically by using the BiLSTM and the CRF model was used to restrict the rationality of labels at the sentence level based on BIOES labeling strategy.The entity identification of grape diseases and insect pests was completed,and the triad data was extracted according to the label categories.Neo4 j was used to store and display the knowledge graph of grape diseases and pests.By adjusting the parameters of the model,the F1 value is improved by 7.86% and 7.82% respectively,compared with the HMM and the CRF,reaching84.69%.The accurate extraction rate of triad data of grape diseases and pests was 89.87%.(2)Design of automatic question and answer method for grape diseases and insect pests based on knowledge graph.An automatic question and answer method of grape diseases and pests based on deep learning was projected to overcome the problem of semantic deviation of questions caused by inaccurate word segmentation results in traditional question and answer methods.The question short text contains few characteristics,various forms of expression and specific problems in the grape field.To complete the comprehension of the intention of the question,the word2 vec was used to represent the character level feature vector of grape disease and insect pests’ text.And the CNN was applied to abstract the question specific.The BiLSTM-CRF is applied to identify the candidate entity of question,and cosine similarity calculation is used to complete the question entity recognition.Cypher statement is used to query the knowledge graph to realize the answer hunt.The research suggests that the F1 of the question intention automatic understanding method reaches 99.92%.The question entity recognition recall rate is 77.86%.The answer query accuracy rate is 64.58%.(3)Design and implementation of an automatic question and answer system for grape diseases and insect pests based on the knowledge graph.The system is based on the knowledge graph of grape diseases and insect pests.Django is used as the development architecture.The deep learning model is applied to achieve the entity identification of grape diseases and insect pests and the knowledge question and answer function,which saves the time for users to obtain information and advances the intellectualization of the grape farming.