Research On Fragmented Agricultural Knowledge Navigation System Based On Label Distribution Learning

With the increasing degree of informationization of the Internet,information retrieval has become the most important way for people to acquire knowledge.Traditional search engines are mostly based on keyword matching.In 2012,the introduction of knowledge graph changed the traditional search engine model.Search engines can intelligently"understand" the natural language input by users and feed back the most relevant knowledge to users.Although scholars have constructed corresponding knowledge graph in many fields,the knowledge graph in the agricultural field is less constructed,the related knowledge is trivial,and the corresponding marks are lacking,so there is no effective organizational form.Therefore,how to reasonably organize domain knowledge to build domain knowledge base,improve retrieval efficiency,and provide high-quality services to users in related fields has become an urgent problem to be solved.In view of the above problems,this thesis uses the combination of knowledge graph and xml topic map to organize domain knowledge,and uses label distribution learning to effectively reduce the search scope,which can effectively organize fragmented agricultural knowledge and achieve good retrieval results.The main contents of this article are as follows:(1)Research on semantic retrieval model based on knowledge graph.Analyze the advantages and disadvantages and similarities of knowledge graph and xml topic maps,using open source knowledge graph as data sources,supplemented by databases such as crop pests and diseases as resources,combining knowledge maps with xml topic maps,and storing agricultural knowledge graph using Neo4j map database.Study the techniques of semantic similarity and semantic extension to realize the semantic retrieval function of knowledge graph.(2)Research on knowledge navigation model based on label distribution learning.Firstly,the Kmeans-based label distribution learning is studied and applied to the knowledge graph semantic retrieval model.Based on the Baye-KM marker distribution learning,the idea of probability and statistics is used after clustering to avoid the input.When the transformed space vector is sparse,it can effectively improve the recall rate of the search,so that the accuracy and recall rate of the search are both good.(3)According to the requirements of knowledge navigation system,the knowledge navigation model based on label distribution learning is designed and implemented,and the corresponding knowledge navigation prototype system design is completed.