Semantic Alignment of Official and Collaborative Geospatial Data: A Case Study in Brazil

Abstract

Geospatial data is crucial for sustainable development, but obtaining up-to-date and high-quality data is challenging in many regions, including Brazil. Collaborative mapping on platforms such as OpenStreetMap (OSM) has produced updated and open geospatial data, especially in urban areas, but its quality is heterogeneous. In addition, semantic interoperability is challenging when integrating OSM data with authoritative geospatial data. This article presents a procedure for semantic alignment between two conceptual models within a conflation process to elicit background knowledge for geospatial data integration. The first model is the Technical Specification for Structuring Vector Geospatial Data (ET-EDGV 3.0) in Brazilian Portuguese, and the second is the OSM model with tags mainly in English. The alignment produced a table combining the ET-EDGV classes, attributes, domains, and geometries with the OSM tags and elements. The semantic alignment was tested in two study areas to check the thematic accuracy of transportation data imported from OSM compared to the data in the reference database. The study found that the best percentage of segments correctly classified by alignment was for "highway=trunk" tags (98.27%) and "highway=primary" (98.20%), corresponding to road and highway segments, and for the "highway=residential" tag (76.20%), corresponding to sections of residential streets. The study also identified factors that may contribute to low accuracy rates, including ambiguous semantic descriptions and the need for local context analysis. This research contributes to adding collaborative data to the official mapping, a relevant alternative for updating and supplementing reference mapping that can be applied in other geographical contexts.