Geographic data plays an essential role in various Web, Semantic Web and machine learning applications. OpenStreetMap and knowledge graphs are critical complementary sources of geographic data on the Web. However, data veracity, the lack of integration of geographic and semantic characteristics, and incomplete representations substantially limit the data utility. Verification, enrichment and semantic representation are essential for making geographic data accessible for the Semantic Web and machine learning. This article describes recent approaches we developed to tackle these challenges.