Geocontext extraction methods analysis for determining the new approach to automatic semantic places recognition

Goal of this paper is to determine actual trends in geocontext extraction methods and to understand which types of geocontext information are the most interesting for users. For this purposes comparison of recent researches about geocontext analysis was done. Researches were compared by the type of achieved result, used formalism, source data and limitations. As the main result of comparison new approach for automatic semantic places recognition was proposed. This approach is based on geotags markup with semantic user-defined tags. The solution allows extracting information (coordinates and a set of corresponding semantic tags on the natural language) about locations which are interesting for the location-based services users. The main advantage of the approach is its simplicity - the method does not rely on any syntax analysis algorithms during the semantic labeling stage. For illustrating the approach an example of the general purpose accidents monitoring service for the Geo2Tag platform was described.