Research On Intelligent Question Answering Method Of Agricultural Diseases And Pests Based On Knowledge Graph

As one of the main agricultural disasters in China,agricultural pests and diseases have the characteristics of diverse types,a large influence range and fast outbreak speeds.The range and severity directly affect the quality and yield of crops.Therefore,in the process of crop planting,agricultural practitioners need to quickly and accurately obtain knowledge of pests and diseases to control them in advance.However,with the development of agricultural information technology,the network data is growing exponentially,resulting in the overall heterogeneous,disordered multi-source state of this data.This makes it difficult for farmers,agricultural technology experts and other personnel to quickly obtain the required information.The intelligent question-answering system can accurately understand the user’s intention and return the corresponding answer in time,which can provide knowledge of diseases and pests for agricultural personnel more conveniently and efficiently.In addition,the knowledge of pests and diseases is stored in the structured form of a knowledge graph,which provides a highquality knowledge base for the question-answering system and helps to mine the correlation between the knowledge of pests and diseases.Combined with a knowledge graph and intelligent question-answering technology,the thesis studies the intelligent question-answering method of agricultural pests and diseases based on a knowledge graph,and designs and implements the knowledge question-answering system on this basis.The main research contents are as follows:(1)Construction of a knowledge graph of agricultural pests and diseases.In view of the lack of a large-scale public knowledge base on pests and diseases in the agricultural field.A top-down method was used to construct the knowledge graph of pests and diseases.In the process of extracting unstructured data,the thesis proposes an entity recognition method for pests and diseases based on radical and character features.Aiming at the characteristics of radical features of pests and diseases,multiple convolution blocks are designed to obtain radical shape information to enhance the text features of pests and diseases.In order to solve the polysemy of Chinese text,a BERT model is used to extract character text information.Finally,the Bi GRU-CRF model is used to extract the text context information and add the rationality constraints to the predicted label sequence.Experimental results show that the accuracy and F1 of the proposed method are as high as 97.23% and 97.82%,respectively.Compared with the mainstream model BERT-Bi LSTM-CRF,it is improved by 2.68% and 2.64%,respectively.(2)Complex question-answering method for agricultural pests and diseases based on knowledge graph embedding.Based on the constructed knowledge graph of agricultural pests and diseases,an automatic question-answering method based on knowledge graph embedding was proposed by taking advantage of the rich semantic and structural features of the knowledge graph.Firstly,Compl Ex was used to embed the knowledge graph into a low-dimensional dense vector space,and dual-induced regularization was introduced to alleviate the overfitting problem.Then,aiming at the problems of the uneven education level of users in the agricultural field and poor standardization of questions,it is proposed to introduce text pinyin features through a text convolutional network to enhance the semantic understanding ability of questions.Finally,the linear model is used to map the question to the vector space of the knowledge graph,and the best answer is screened out by the scoring function.Experimental results show that the average F1 value of the question-answering method proposed in the thesis is 87.42%,which is21.68%,12.14% and 9.59% higher than that of Key-Value,Embed KGQA,and MHCKBQA models,respectively,and the effect is significant.(3)Design and develop a question-answering prototype system for agricultural pests and diseases based on a knowledge graph.The constructed knowledge graph of agricultural pests and diseases is used as the knowledge base.The question-answering method model embedded in the knowledge graph is used as the core processing module,and realizes the front-end interaction of the question-answering system combined with Django,Python and other technologies to complete the answers to the information such as damage symptoms,control methods and damage parts in agricultural pests and diseases.