Design And Implementation Of Fertilizer Knowledge Question Answering System Based On Knowledge Graph

China is the world’s largest consumer of fertilizers,and fertilizers play a vital role in agricultural production.It is an essential task to study how farmers can have scientific and effective knowledge of fertilizers such as fertilizer distribution and application.Intelligent knowledge Q&A system can provide farmers with professional knowledge services such as fertilizer dispensing and fertilizer application,but the existing intelligent knowledge Q&A systems all store knowledge information in the form of traditional databases,which has a low knowledge utilization rate.Knowledge graph stores and associates knowledge information in a structured way to maximize the use of knowledge information,Designing and implementing a knowledge graph-based fertilizer knowledge Q&A system can provide farmers with more efficient and accurate Q&A services and a better user experience.This paper obtains fertilizer knowledge text data from fertilizer companies and agricultural websites such as China Fertilizer Network.The intelligent question and answer method based on knowledge graph is studied,and to design and implement a fertilizer knowledge question and answer system based on the knowledge graph.The main studies are as follows:(1)A named entity recognition model based on BERT with adversarial training is proposed to address the problem of inaccurate recognition of entities with fuzzy fertilizer knowledge domain boundaries,which affects the accuracy of model entity recognition.The perturbation factors are generated by introducing the Free LB adversarial training method,and combined with the vectors of the word embedding layer in BERT to form the adversarial samples to improve the model’s recognition ability for boundary ambiguous entities by adversarial training.The experimental results show that the model proposed in this paper improves the fertilizer knowledge entity recognition compared with named entity recognition models such as BERTBi LSTM-CRF.(2)Knowledge graph-based knowledge question and answer method,the problem that the fertilizer question contains less feature information,which affects the accuracy of question intention recognition,the commonly used question intention recognition models BERT,Text CNN and BERT-Text CNN are compared and experimented,the BERT-Text CNN model with the best effect is selected as the interrogative intent recognition model in this paper and applied to the fertilizer knowledge question and answer function in the question and answer system.(3)Design and implementation of a knowledge graph-based fertilizer knowledge question and answer system.The system includes fertilizer knowledge question and answer function,fertilizer industry information function,fertilizer expert team function and system user management function.The fertilizer knowledge question and answer function provides users with fertilizer knowledge question and answer services.Fertilizer product information function provides users with detailed information on fertilizer products for their reference.The Fertilizer Expert Team feature provides information and contacts about fertilizer experts,facilitating communication between users and experts.