February 09, 2015
Fujitsu Research and Development Center today announced a Chinese enterprise knowledge graph construction based on Linked Data technologies. First, cross-lingual schema matching is used to map enterprise data schema and standard metadata schema. Then, knowledge validation technologies are adopted to data conflict resolution and data fusion. Finally, enterprise knowledge and web knowledge are linked by entity linking technologies.
Compared with conventional technologies which only integrates monolingual enterprise data at data layer, our technology can integrate multilingual data at both data and semantic layer using Linked Data technologies. Moreover, our technology can integrate enterprise data with external knowledge base. Our knowledge graph construction technology will provide enterprises with more efficient and intelligent data management solution and richer knowledge for enterprise decision.
【 Background 】
Enterprise Knowledge Graph is a crucial infrastructure to structure, integrate and manage the knowledge of organizations. However, constructing enterprise knowledge graph faces several challenges such as heterogeneous and inconsistent data, lack of semantic description and lack of links with external knowledge base. We developed a novel enterprise knowledge graph construction technology including cross-lingual schema matching, knowledge validation and cross-lingual knowledge association.
【 Topics 】
Conventional technologies mainly focus on integrating data distributed in multiple databases and usually integrate monolingual enterprise data at data layer. They cannot use standard metadata schema to describe the data, therefore enterprise data cannot be associated with web knowledge which will hinder enterprises using external knowledge to make decisions. Meanwhile, conventional technologies cannot solve data conflict and cross-lingual issues efficiently. How to make enterprise data and web knowledge association and enterprise data schema with standard metadata schema mapping is very important for the enterprise knowledge graph construction method.
【 Technology 】
Enterprise knowledge graph construction based on Linked Data includes three key technologies, cross-lingual schema matching, knowledge validation and web knowledge association. Figure 1 illustrates the cross-lingual schema matching technology. This technology can map enterprise data schema with standard metadata schema. The main benefits are: (1) Enterprise data could inherit semantic information from standard schema. (2) Enterprise data could be efficiently integrated and reused by other applications.
Figure 1: Cross-lingual schema matching illustration
Figure 2 shows the knowledge validation technology. When knowledge conflict occurs among data sources A, B and C, knowledge validation technology exploits data source authority, data redundancy and consistent and semantic constraints to identify the most reliable data value and then integrate the data sources.
Figure 2:Knowledge validation illustration
Figure 3 shows the knowledge association technology including internal knowledge association and external knowledge association such as DBPedia and GeoNames. After knowledge association, enterprise could take advantage of richer knowledge for decision making.
Figure 3: knowledge association illustration
【 Future Plan 】
Fujitsu R&D Center will popularize our image stitching technology to libraries for large ancient document digitization with our partners. We will improve our technology based on the feedbacks of customers.