How to Ask Maps Questions

Until very recently, historians have not accessed digital collections of map scans with the intention of doing more than visually browsing them, or using individual sheets to illustrate a point developed with evidence from other primary sources. MapReader is a computer vision pipeline developed within the Living with Machines project which was designed to help historians ask questions of large collections of digitized series maps. Using this software library means that we can transition from informal insights about how different features appear on a small number of maps to formal, structured data about where they are across the whole nation. MapReader enables computational research with maps, changing the way that historians can interact with cartographic information about the built and natural environments of the past. Using examples from experiments with about 16,000 nineteenth-century Ordnance Survey maps digitized by the National Library of Scotland, this presentation examines, first, how the collaboratively-developed MapReader creates datasets documenting the location of expert-defined map features across thousands of maps and, second, approaches to analyzing this data given that the results of the pipeline contain errors.