Research And Application Of Construction Method Of Rice Knowledge Graph In Ningxia

Since Google proposed the concept of knowledge graphs in 2012,many scholars have continuously explored the representation of knowledge graphs of complex data across domains.Knowledge graph is an application-driven knowledge representation.It is more suitable for representing domain knowledge with complex structure and complicated relationship than traditional representation method.At present,the knowledge representation methods for agriculture mainly use vocabulary,RDF and ontology.With the development of knowledge graph technology,some research results of agricultural knowledge graphs have been produced.However,many research results,especially in the field of rice Knowledge representation is mainly focused on the definition of the concept layer structure.Based on the knowledge representation model,this study is aimed at the agricultural field,combined with the application needs of the rice industry in Ningxia,focusing on the knowledge graph representation method of rice data in Ningxia.Starting from the acquisition and processing of rice data,the crawler program is developed to obtain the rice data on the web and public datasets and the agricultural semantic dictionary,and comprehensively apply various natural language processing techniques to complete data preprocessing;call D2 R to convert structured data into RDF.Schema;for semi-structured and unstructured data,the semantics and regular expressions are used to extract the entities and their attribute values,and the pattern matching method is used to extract the non-categorical relationships;the extracted multiple RDFs are imported into Protégé In the process of integrating ontology;under the guidance and guidance of rice experts,the rice knowledge graph is constructed and improved.The research content of this paper mainly includes :(1)Data acquisition is mainly carried out from the data center of rice related institutions,agricultural books and web-side data.The data of rice field is obtained by using various techniques such as scrapy crawler framework and PDF electronic analysis tool.Call the Pandas,Numpy toolkit and LTP tools in Python to complete the data preprocessing process for denoising,deduplication,cleaning,word segmentation,part-of-speech tagging and dependency syntax analysis of rice field data.(2)Differential knowledge extraction for rice data of different data structures.Structured data is converted to RDF format using D2 R tools;semi-structured and unstructured data is constructed by means of agricultural semantic dictionary,entity and attribute values are obtained by defining rules,and pattern matching method is used to extract non-categorical relations.(3)Combine and complete the rice domain knowledge base by means of knowledge fusion and knowledge reasoning related technologies.The multiple RDF generated by the knowledge extraction is imported into Protégé.With the Merge ontogies in the Refactor option,the ontology fusion mechanism is used to fuse multiple domain ontology,and finally the complete Ningxia rice knowledge graph is formed.On this basis,the SWRL rules are written,and knowledge reasoning and completion are completed under the guidance of domain experts.In order to evaluate the availability of the rice knowledge graph and the validity of the knowledge graph construction method,this paper designs and implements a query platform based on rice knowledge graph to verify its application.