An Automatic Knowledge Graph Creation Framework from Unstructured Text

Abstract

Knowledge Graph creation from unstructured text plays a crucial role in the semantic web community. Consequently, there are many approaches proposing to create Knowledge Graph from unstructured text. However, Knowledge Graph integration is omitted. Knowledge Graph integration is an essential procedure because it could reduce the heterogeneous problem and could increase searchability over Knowledge Graphs. In our previous work, we proposed the T2KG framework, an automatic framework for Knowledge Graph creation from unstructured text, with keeping the integration issue in mind. Although we could achieve better results to create a Knowledge Graph than the previous approaches, the reasonable precision is still not reached. In this paper, we therefore propose T2KG_Ext: an extension of the T2KG framework. In the T2KG_Ext framework, we re-organize the T2KG framework to increase the ability to generate candidate triples and introduce the extension component, namely candidate selection, to the T2KG framework. In the preliminarily experiments, we reported the problem of the T2KG framework and showed some evidences that the T2KG_Ext could deal with such problems.