Ambiguity and plausibility: managing classification quality in volunteered geographic information

Abstract

With the ubiquity of technology and tools, current Volunteered Geographic Information (VGI) projects allow the public to contribute, maintain, and use geo-spatial data. One of the most prominent and successful VGI project is OpenStreetMap (OSM), where more than one million volunteers collected and contributed data that is obtainable for everybody. However, this kind of contribution mechanism is usually associated with data quality issues, e.g., geographic entities such as gardens or parks can be assigned with inappropriate classification by volunteers. Based on the observation that geographic features usually inherit certain properties and characteristics, we propose a novel classification-based approach allowing the identification of entities with inappropriate classification. We use the rich data set of OSM to analyze the properties of geographic entities with respect to their implicit characteristics in order to develop classifiers based on them. Our developed classifiers show high detection accuracies. However, due to the absence of proper training data we additionally performed a user study to verify our findings by means of intra-user-agreement. The results of our study support the detections of our classifiers and show that our classification-based approaches can be a valuable tool for managing and improving VGI data.