Applying Commercial Computer Vision Tools to Cope with Uncertainties in a Citizen-driven Archive: The case study Topothek@exploreAT!

Abstract

Uncertainties in data, e.g., incomplete data sets, data quality issues or inconsistencies in annotations, are a common phenomenon across disciplines. How to address these issues is context dependent. In this paper, we address uncertainties in the citizen-driven archive Topotheque as a concrete use-case in the Digital Humanities project exploreAT!, and demonstrate, how to deal with uncertainties by benchmarking a set of selected commercial computer vision (CV) tools. The approach aims to enrich Topotheque's data to enable better access, connectivity and analysis for both researchers and citizens. Results show that by applying CV, existing uncertainties are noticeably reduced, but new ones also introduced. Better grounds for semantic structuring are provided, enabling higher connectivity and linking within Topotheque, but also across other data sets. Ultimately, the enrichment of the archive is for the benefit of both researchers and citizens enabled by addressing and tackling apparent uncertainties.