Spatiotemporal features based geographical knowledge graph construction
ZHANG Xueying,ZHANG Chunju,WU Mingguang,LV Guonian
DOI: https://doi.org/10.1360/ssi-2019-0269
2020-01-01
Abstract:Geographical knowledge is the cognitive result of human beings on the spatial distribution, evolution, and interaction of geographical phenomena or things. At present, there is a common phenomenon, i.e., “mass of data, information explosion, and hard to find knowledge” in geographic information services in the context of big data. The geographic knowledge graph (GeoKG) is a knowledge-based system that uses the semantic network to describe geographic concepts, entities, and their relations. It has application prospects in geographic knowledge understanding, geological problem solution, as well as spatiotemporal prediction and decision-making. In addition to the characteristics of general knowledge, geographic knowledge also has specific characteristics of spatiotemporal and geoscience mechanism. Therefore, the construction and applications of GeoKGs have both generality and certain professional specialty. Based on the spatiotemporal characteristics of geographic knowledge and the structure of knowledge graphs, we propose a method for GeoKG construction that considers the characteristics of time and space. Firstly, we describe the basic ideas and technical processes of GeoKG construction and briefly explain the main research contents of geographic knowledge acquisition, geographic knowledge abstraction and representation, as well as geographic knowledge organization and management. Secondly, based on the analysis of basic questions answered by geography, we summarize and abstract geographic knowledge elements and then explore a geographic knowledge representation model with three levels of “geographical concepts-geographic entities-geographical relations to describe semantic units of geographical knowledge and their logic relations. Finally, the spatiotemporal features are assumed as the basic conditions, and based on the statuses of geographic entities, we propose a formal method based on “process-relation to represent the evolution process of geographic entities and complex geographic relations. In conclusion, our results can effectively solve the basic problems of GeoKG construction, which are crucial for the acquisition, fusion, reasoning, and application of geographic knowledge. Furthermore, it can be extended in geoscience fields such as geology, environmental science, and meteorology, and has the prospects for the promotion of geoscience knowledge services.