Popis: |
In the era of Web of data and Social media, the recent concept of enterprise knowledge graphs (EKG) emerges as a backbone for federating valuable web open information along with the data contained in internal corporate databases. This work proposes KGMap a mapping schema for leveraging knowledge graphs by bridging between heterogeneous Relational, Social and Linked Data Web sources. The KGMap approach connects the knowledge graph schema with the sources metadata elements relying on semantic similarity measures. The implementation of the proposed approach is a configurable middleware for generating enterprise knowledge graphs. We assessed the middleware in a commercial enterprise use case where inputs are integrated on-the-fly from an internal CRM database combined with DBpedia SPARQL endpoint and Facebook Web API. We conducted an empirical study to test the effectiveness of KGMap using different similarity measures. Furthermore, we compared the KGMap with state-of-the-art approaches. The evaluation results illustrate a better precision and recall of the contributed approach, which allowed the data to be more accurately integrated into a knowledge graph. |