The Construction And Application Of Tobacco Redrying Knowledge Graph

Tobacco redrying,as an important part in the processing of tobacco production,generates a large number of relevant knowledge data every year.However,due to the lack of effective methods to sort out the tobacco redrying knowledge systematically,the tobacco-redrying knowledge becomes numerous,disordered and scattered.Thus tobacco redrying enterprises can not get the necessary knowledge from huge amounts of data to provide strong support for operation and management.In order to make effective use of the knowledge data in the field of tobacco redrying,this thesis integrates the redrying data with the powerful knowledge interconnection organization ability of the knowledge graph and focuses on the knowledge extraction technology in the knowledge graph.In this way,this thesis constructs the knowledge graph based on tobacco redrying.Meanwhile,the tobacco redrying knowledge graph system is designed and built,which intuitively displays the obtained redrying knowledge in the way of graph structure,and realizes the intelligent applications of knowledge graph.The main research work of this thesis is as follows:(1)On the basis of Bi-LSTM-CRF model,an improved model of tobacco redrying named entity recognition is constructed in this thesis.First of all,the model applies the BERT model to study the feature representation of sentences from the input text.Then uses the Bi-LSTM model to get the context feature information of words.Finally,the CRF model is applied to get the optimal entity tags at the sentence level.The comparison experiments with other traditional named entity recognition models show that the BERTBi-LSTM-CRF constructed in this thesis has achieved a good effect on the named entity recognition task in the field of tobacco redrying.The accuracy,recall and F1 of the model are achieved 89.45%,88.91%? 89.21%.(2)This thesis constructs the relationship extraction model of tobacco redrying based on Bert-Bi-GRU-Atten-CRF.Combined with the advantages of BERT model in feature acquisition,this model uses Bi-GRU model with less parameters and faster model calculation speed to obtain the context relationship between entities and relationships in the input sentence,and then reallocates the word weight through the attention mechanism.Finally,CRF is used to comprehensively consider the constraint relationship in the sentence to get the globally optimal entity relationship label.Through the comparative experiment with the traditional relationship extraction model,the Bert-Bi-GRU-Atten CRF relationship extraction model constructed in this paper is effective in the tobacco redrying domain relationship extraction task,and the precision,recall and f1 score of the model evaluation index have reached 86.64%,87.88% and 86.69%.(3)A tobacco redrying knowledge graph system is designed and implemented.Based on tobacco redrying named entity recognition and relationship extraction,this thesis constructs the tobacco redrying knowledge graph.At the same time,combined with tobacco redrying knowledge graph,this thesis completes the design and construction of the corresponding knowledge graph system,which can mainly include named entity recognition,entity query,relationship query and Knowledge Q ? A.