| In recent years,knowledge graph has developed rapidly and become an important research direction of artificial intelligence.It has been widely used in various fields.Domestic and foreign scholar’s have carried out quite in-depth research on each part of knowledge graph construction.Taking agricultural crop knowledge graph as the research object,this paper conducts research on the model construction and attribute extraction of agricultural crop knowledge graph by using the Thesaurus of Agricultural Science,Conditional Random Field Model,Bi-directional Long Short-Term Memory model and BERT model.The main research work of this paper is summarized as follows:(1)This paper analyzes the research background and current situation of crop knowledge graph,researches related technologies and completes the construction of crop knowledge graph.(2)Most of the existing knowledge graph data sources are public resources such as wikipedia or Baidu encyclopedia.In view of the shortage of knowledge resources in the field of agriculture,the construction method of crop knowledge graph was studied by using the knowledge graph construction technology based on thesaurus.On the basis of the preliminary crop knowledge graph constructed from thesaurus to knowledge graph pattern layer and data layer,the attributes of each crop are added.(3)At present,in the field of knowledge graph,the existing research methods of attribute extraction are all public data sets,which have little application in the field of agriculture.Aiming at the improvement of the existing bilstm-crf model,a BERT-bilstm-crf model based on BERT pretraining is proposed.The data set was divided into training set and test set,and the ratio was randomly assigned to 3:1.Compared with the bilstm-crf model,the accuracy of the bert-bilstm-crf model is 3.58% higher than before,the recall rate is 2.17%higher,and the F1 value is 2.75% higher,which meets the requirement of attribute extraction in the field of agricultural crops. |
PDF Full Text Request