UC Santa Barbara

Geo-ontologies play significant roles in formalizing concepts and relationships in geography as well as in fostering publication, retrieval, reuse, and integration of geographic data within and across domains. The status quo of geo-ontology engineering is that a group of domain experts collaboratively formalize the key concepts and their relationships. On one hand this centralized top-down ontology engineering approach can take into account invaluable expert knowledge and capture our perception of the world correctly in most cases; on the other it might yield biased geo-ontologies and misrepresent some important concepts or the interplay between different concepts due to the fact that such top-down ontology engineering strategy hardly takes into consideration the existing dataset. With an increasing number of Linked Data on the Web, we are able to use such data to assist the traditional geo-ontology engineering process. However, the quality of Linked Data also imposes challenges to this task. This research proposes a framework by modeling the hierarchical structure using a series of data mining algorithms and eventually quantifies the difference between the original ontology and the data mining one with the proposed Discrepancy Index. The Discrepancy Index can help geo-ontology engineers identify as well as quantify potential ontological modeling issues and Linked Data quality issues, thus closing the gap in the dynamic process of geo-ontology engineering.