Research On Grain Intelligent Question Answering System Based On Knowledge Graph

With the gradual development of question answering systems,question answering systems in professional fields can combine knowledge in multiple professional fields to form question answering systems,and their practicality has aroused the interest of professionals.However,there is currently a lack of public crop knowledge graphs in China,and there is a big problem in how to build better quality knowledge graphs in the crop field.There are also certain difficulties in the development of intelligent question answering systems based on deep learning,because the question answering systems based on knowledge graphs require a strong understanding of natural questions,and the agricultural field lacks question answering training corpora based on knowledge graphs.In response to the above problems,this paper integrates deep learning and rule matching,and basically completes the research on the grain question answering system based on knowledge graph.The research content is also divided into the following four aspects:(1)Acquisition technology of structured data: various semi-structured data and unstructured data related to grain varieties are obtained from domestic websites such as China Agricultural Information Network,Agricultural Agricultural Blog,Wannong.com,and Baidu Encyclopedia.data,and semi-structured data with better quality can be directly obtained from the project website of "Shanghai Agricultural and Rural Big Data Sharing Service Platform Construction and Application" on the National Natural Science Foundation of China,which can be further used for the model construction of this subject data.At the same time,a large number of data annotations have been made,and these data have certain contributions to the development of agricultural artificial intelligence(AI).(2)Model construction: For the construction of the model,in the knowledge extraction chapter of this topic,according to the characteristics of the long and many fields to be extracted in the grain text,the targeted solutions are expounded.The data extraction task is carried out through four directions: text classification,named entity recognition,relationship extraction,and rule extraction.In addition,the self-developed self-feedback deep learning model has achieved good results in grain data extraction.The F1 value is now On the test set,it reaches 0.7254,which is a good improvement compared to the public test set at this stage,and the improvement value is 0.0194.(3)Data storage: Because some grain data is updated quickly,and the structure is simple and does not require in-depth reasoning,this topic not only stores the triple data extracted by the model into the knowledge map,but also designs My SQL.Data storage table to answer related knowledge.(4)Smart Q&A: Based on the data information of China Agricultural Information Network,Nongbo.com,Wannong.com,and Baidu Encyclopedia,this project provides grain price trends(selling prices,futures fluctuations,etc.),basic information,irrigation Guidance,sowing guidance,pest prevention,trade,production,cost-benefit,climate type,etc.,are expanded in content compared to the common question-and-answer system that only answers questions about crop diseases and insect pests.This paper proposes a self-feedback algorithm model of the grain intelligent question answering system based on deep learning and rule matching.The experimental results show that the method can solve the problems existing in the traditional manual questioning method to a certain extent,and can effectively improve the correct rate and efficiency of answering.