Research On The Construction Method And Application Of Local Customs And Human Conditions Knowledge Graph Based On Multi-source And Heterogeneous Data

With the rapid development of informatization and big data era,the information that people face in life is becoming more and more complicated,but the information that can really help is very limited.The reason is that the information is not targeted and standardized.Sex.An effective solution is to structure the information in a targeted manner.Knowledge Graph has received more and more attention because of its advantages in data relevance and structure.At present,the knowledge graph has been applied in many fields,including open fields,as well as vertical fields such as medical care,education,and movies.Local customs are the summation of the unique natural environment,customs,etiquette and habits of a place,which has high humanistic and natural value.However,the construction and application of knowledge graphs in this field is still in the preliminary exploration stage,so it is necessary to study the construction methods and applications of local customs knowledge graphs.There are not enough data sources for traditional knowledge graph construction methods.For extracting entity relationship triples from text,traditional methods are not ideal for extracting triples with overlapping entities.In view of the above situation,this article provides a solution to the construction and application of knowledge graphs around the field of local customs and conditions.The main research contents are as follows:(1)Taking Inner Mongolia as an example,construct a knowledge map of Inner Mongolia’s local customs.For the field of local customs,some optimizations are made on the basis of several traditional vertical domain ontology construction methods to make the ontology construction method more suitable for the field of local customs.Under the constraints of ontology,based on data from multiple sources and different structures,using information extraction technology,entity relationship triples are extracted from them,and stored in the graph database after knowledge fusion,to complete the preliminary construction of the Inner Mongolian customs knowledge map.(2)Using a hierarchical annotation method for reference,a hierarchical annotation model based on BERT-CRF is proposed to jointly extract entities and relationships from the text.First,BERT is introduced to encode the input text,and CRF is used to mark the subject of the triple.On this basis,for each predefined relationship,the half-pointer and half-marking method is adopted,and the double pointer is used to mark the beginning and end positions of the corresponding subject and the object under the relationship in the text,and then the triples are completely extracted.Compared with the original hierarchical labeling and the traditional triple extraction method,the triple extraction effect has been improved.(3)Based on a pipelined knowledge graph question-and-answer method,the entity mention recognition module is improved,and the object extraction idea in the hierarchical labeling of triples is used for reference,and a half-pointer and half-labeled entity extraction method is proposed.And the recognition method,using double pointers to mark entity mentions from the question sentence,the recognition effect and the question and answer effect are improved.(4)Based on the Inner Mongolia local customs knowledge map and knowledge map question-and-answer method,design and implement the Inner Mongolia local customs knowledge question and answer platform,including the question and answer function based on the knowledge map,the visualization of the knowledge map and the display function of the entity picture,and allow users to submit triples to Update the knowledge graph by crowdsourcing.