Ground Truth and the Humanities

--- tags: discussion, gazetteer --- # Ground Truth and the Humanities ## The Structured Representation of Places in Knowledge Bases In the worlds of map making and remote sensing, [ground truth](https://en.wikipedia.org/wiki/Ground_truth) is information collected from direct observation of land that can be used to [calibrate](https://placesjournal.org/article/skywatching/?cn-reloaded=1) and [verify](http://fieldpapers.org/) data collected from other sources (e.g. satellite imagery). What counts as 'ground truth' about historical places? ('Authority records' from a national library? Wikidata? Geonames? A scholarly gazetteer?) Why? Who decides? Is this even a useful concept in the humanities? #### If this is enough to spark your interest in a discussion, please join us on the 1st of December at 4pm London time. #### Read on to find some more digital humanities context to mull over. Place is a useful common denominator in humanities research. It is a unit of analysis that traverses primary sources and the data we create about them (e.g. GLAM records). Researchers and curators grapple with a) the instability of place names (aka toponyms) over time and between cultures, and b) the relationships between instances of place names in sources, the concept of a place, and their geospatial locations. These challenges have inspired projects and tools like Recogito, the World Historical Gazetteer, the EMPlaces data model, and many more around the world. In turn, these projects depend on theories of place that connect human expressions of ideas preserved in text (e.g. the string 'San Francisco') to machine-readable information about location. Many assumptions about the connection between a place name and that location are embedded in the strategies that we use to perform this linking (e.g. manual or automatic geoparsing). - How explicit do we make these assumptions? What are their implications? - Can there actually be a ground truth spanning across time? --> Check out some thinking about this from the sadly defunct [Who's On First gazetteer project](https://www.whosonfirst.org/blog/2016/08/15/mapping-with-bias/) - Does the process of building gazetteers or annotating texts to generate gold standard datasets for Named Entity Recognition (of place entities) mimic the process of establishing ground truth? - **What is the relationship between 'authority records' (e.g. Wikidata or the other knowledge bases that we use as 'spinal gazetteers') and this idea of 'ground truth'?** - As we build knowledge bases collecting and linking information about historical place, is 'ground truthing' a useful activity to verify and calibrate historical data? How would we do that? - How do these decisions about place impact the way we handle other kinds of Named Entities (people, organizations, etc.) Let's discuss!