Research On Sheep Disease Diagnosis Method Based On Knowledge Grap

(1)Sheep farming is an important component of China’s livestock industry,and its rapid development has created an urgent need for sheep disease diagnosis and prevention.The core of sheep disease prevention and control lies in accurate diagnosis of diseased sheep,which relies on specialized knowledge in sheep veterinary medicine.However,there is a shortage of sheep disease experts who possess sufficient professional knowledge and experience to effectively diagnose and treat diseased sheep.Additionally,automated sheep disease diagnosis systems suffer from low efficiency and poor robustness,failing to meet the requirements of real-time and accurate diagnosis in the sheep disease diagnostic process,which may result in potential economic losses.Addressing the aforementioned issues,the main research work of this paper is as follows: Constructed a corpus of sheep disease texts.In view of the lack of annotated corpus data resources in the field of sheep disease diagnosis,this paper integrated sheep disease diagnosis data from sheep farms and several sheep farming encyclopedias,proposed the R-BIO annotation method based on the characteristics of these data and combined with the opinions of sheep disease experts.The Chinese sheep disease corpus was annotated using YEDDA,and a140,000-word sheep disease text corpus was annotated by combining dictionary pre-annotation with manual annotation,and the sheep disease text corpus was constructed.(2)Proposed a joint extraction model of entity relationships in the field of sheep disease and constructed a sheep disease knowledge graph.In view of the lack of open-source knowledge graphs in the field of sheep disease,this method first used the BERT language model to encode the sheep disease text corpus,and then realized the recognition of subjects,relationships,and objects through the subject prediction module and the joint prediction module of object relationships.Afterwards,use the trained sheep disease entity-relation joint extraction model to automatically extract information from the raw sheep disease data,achieving the completion of sheep disease knowledge.Bring the F1 value of the model to 74.14%.Finally,the triple information was stored using the Neo4 j graph database to construct the sheep disease knowledge graph and complete the visualization of the sheep disease knowledge graph data.(3)Proposed a sheep disease diagnosis method that integrates knowledge graphs and graph neural networks.This method uses graph neural networks to aggregate structured disease information in the sheep disease knowledge graph and integrates it with feature word vectors in the symptom description text as multi-channel inputs of the long short-term memory network.It learns disease and symptom features from both knowledge and semantics.Make the accuracy and recall rate of sheep disease diagnosis reach 94.29% and 88.26% respectively and finally realizes that when the user inputs specific symptoms,the model can obtain the K diseases with the highest confidence and give the incidence probability of each disease,providing guidance for sheep disease diagnosis.