Dynamic Semantic Modeling of Structural Data Sources

Hailong Chu, Weiming Wu, Wolfgang Mayer, Debo Cheng, Hong-Yu Zhang, Zaiwen Feng

Abstract: A semantic model of a data source is a represen-tation of the entities and relationships contained in the data. Building semantic models is a prerequisite for automatically publishing data to a knowledge graph. However, constructing these semantic models is a complex process requiring consid-erable manual effort and a high potential for errors. Several approaches have been proposed for automatically identifying semantic models of data sources. However, as the schemata of data sources change, the corresponding semantic models may require amendments. We propose a dynamic semantic modeling approach whereby the semantic model of a data source can be adapted automatically as data sources change. We show that our incremental approach outperforms state-of-the-art systems by an average of 7.4% in the F1 score and that it is orders of magnitude faster than constructing a semantic model from scratch.