Research On Semantic Intelligence Generation Technology Of Data Asset Based On Graph Computing

In today’s big data era,enterprises often store massive amounts of data,which is also an asset for enterprises.However,it is difficult for enterprises to efficiently obtain information and exert value from massive and complex data.Therefore,it is very important to organize and understand data through enterprise data asset inventory work.Data standard documents and database systems are usually important data description sources for enterprises.In the inventory process,the data standard document is an important reference for business personnel to understand the enterprise database system.Actually,due to the large number of people implementing the database system and the rapid update of application scenarios,there is often a mismatch between the two.When performing inventory manually,it is necessary to integrate the information of different data description sources to understand the field information,which also means huge manpower and time costs,and requires the inventory personnel to have a certain business understanding.As an efficient structure of semantic interconnection,knowledge graph provides a new idea for processing heterogeneous data of enterprises.Knowledge graph alignment is an important way to integrate information related to equivalent entities in different knowledge sources.In order to solve the problem of incomplete filed business semantic description information under different data description sources in the process of enterprise data asset inventory,a technology for intelligent generation of data asset business semantics based on graph computing is designed.The main work is as follows:(1)The definition of relationship sets for enterprise data standard documents and database systems are proposed,and the corresponding knowledge graph construction methods are designed: the definition of relationship sets utilizes the field pattern information available in data documents and databases,including the information of elements themselves,schema structure information and data instance information,etc.And the corresponding knowledge graph construction methods are proposed to fully model the fields in different data description sources.(2)A multi-view entity alignment model based on graph neural network is proposed:For multi-type relationship of enterprise knowledge graph,two feature processing angles of common-view and specific-view are introduced.The global structure information of entity and different types of relationship information are captured separately,and finally combined for entity alignment,which effectively improves the accuracy of the entity alignment task.(3)A knowledge graph-based intelligent generation method for data asset business semantics is proposed: based on the alignment results of knowledge graphs of different data description sources,an automatic generation method for field business semantic description information based on templates is proposed.In this way,the business semantic description information of fields in different data description sources is completed,helping business personnel to better understand and inventory enterprise data assets.(4)Based on the above work,we designed and implemented a software system of data asset business semantics based on graph computing.The system mainly includes a heterogeneous data processing module,a heterogeneous graph construction module,a heterogeneous graph alignment module,a database interaction and management module,and a business semantic description information generation module,etc.