Kolloquium Geoinformatik Tackling the classification accuracy in Volunteered Geographic Information

Crowd-sourcing, especially in form of Volunteered Geographic Information (VGI) significantly changed the way how geographic data is collected and which products are generated from them. In VGI projects, contributors’ heterogeneity fosters rich data sources, however with problematic quality. In our work, we tackle the quality from Classification perspective. We address two classification problems: Hierarchical classification where the classification of an entity should follow a strict structure and Arbitrary classification where the classification does not follow any strict structure. In VGI projects, classification has some challenges: In some cases, the classification of entities depends on individual conceptualization about the environment. Whereas in other cases, a geographic feature itself might have ambiguous characteristics. These problems lead to inconsistent and inappropriate classifications. To face these challenges, we propose an approach. The approach is mainly based on investigating the geographic context of a specific class of entity. It employs data mining algorithms to develop a classifier. Then, the developed classifier could be used to guide the contributors during the contribution process. The approach acts not only to improve the quality of data, but also it could be used to improve contributors' experience. However, dealing with VGI is tricky and in geographic context there is no universal rule. The empirical studies emphasized the feasibility of the proposed approach, however with some limitations. Currently, towards real evaluations we are working to implement the approach into action phase. During the talk, I would present the ideas beyond the proposed approach and discuss another hot issues related to the quality of VGI.