A Comparative Study of TransE and TransH Algorithmsin Spatial Address Representation Learning

Geographic knowledge graph is the extension of knowledge graph in geography. As the basic data of geographic data, spatial address contains rich spatial information and semantic information. It is a simplified geographic knowledge graph. In this paper, we integrate geographic knowledge into spatial addresses, study how to combine spatial information with semantic information by knowledge representation learning. The spatial address data set is trained on the TransE and TransH, two classical translation models, and a comparative study is conducted through triple classification and distance estimation between vectors.