The David Rumsey Map Collection has been home to tens of thousands of historical maps, and now you can search the collection by the text in the maps instead of just through metadata:
About 57,000 of the georeferenced maps from the collection have been processed with a machine learning tool called mapKurator to collect all the text on maps, i.e. every piece of text (printed or handwritten) that appears on a map, including place names, but also information like a map’s title or its scale, or the names of the people involved in producing the map. Through this process of automatic annotation, this approach turns the text on maps into structured data (text on maps data): the enormous amount of data that has been collected from these maps is now searchable.
