[02/12/2020] Ground Truth and the Humanities: The Structured Representation of Places [and other Named Entities] in a Knowledge Base

[fedenanni]

Idea of a discussion group on knowledge representation and especially how the concept of ground truth is complex when speaking, for instance, about places / coordinates / maps

[kmcdono2]

Hooray! When are you thinking this would get scheduled? Are there other topics in the line up?

Kaspar and I were just talking about this in the context of his annotation article, and I remembered the first time I ever heard someone say 'ground truth' - it was to talk about this: http://fieldpapers.org/ (in the context of prepping for a course on the history of mapping). I was totally blown away by the concept of ground truth and how the US national mapping agency now no longer makes maps based on ground surveys - they rely on volunteer data updates. So there is this broad concern in the mapping community that the growing disconnect between survey and map is the source of all kinds of problems: e.g. inequality in all things that depend on up-to-date, high-res spatial information. This raises some questions I'm interested in:

• when did geospatial data become attached to, and then divorced from, the map?
• where does the phrase 'ground truth' come from and why?
• what are the attributes of ground truth?
• who decides what counts as ground truth and how does this vary by discipline/application? - what does it mean to talk about ground truth when the thing you are interested in is how space changes over time?
• how do you represent this all with data structures that do not displace complexity?

And, finally, how does discussion about ground truth in map world relate to the use of this term (alongside 'gold standard') in data science/ML? Assertions about the 'truth' of something vs. inferred information about that same thing may have started when literally talking about the representation of the ground, but now this has a bigger life in ML.

Some useful readings:

• https://en.wikipedia.org/wiki/Ground_truth
• https://www.whosonfirst.org/blog/2016/08/15/mapping-with-bias/ (we might be able to track down Aaron, the lead developer)
• chapter about historical place modeling by Arno Bosse, begins p79 in this: https://univerlag.uni-goettingen.de/bitstream/handle/3/isbn-978-3-86395-403-1/cost_hotson.pdf?sequence=6&

And...something fun:

• https://placesjournal.org/article/skywatching/

[fedenanni]

@kmcdono2 this is just awesome! Do we want to put this together for a discussion in either Oct or Nov? We don't have anything scheduled at the moment ;-)

[kmcdono2]

Nov?

[fedenanni]

@kmcdono2 would Nov 11th 4pm work? (so much looking forward to this!)

[kmcdono2]

@fedenanni With apologies for the delay...

Text for email over here

It could do with some editing, but it's getting there.

To invite: Leif Isaksen, Elton Barker, Rainer Simoon, Valeria Vitale, Rebecca Kahn, Arno Bosse, Howard Hotson, Miranda Lewis, Nicole Coleman, Karl Grossner, Ruth Mostern, Patricia Murietta-Flores, Rombert Stapel (KNAW), Ian Gregory, Humphrey Southall, Paula Aucott, Jim Clifford, Ali Yaycioglu, Janelle Jenstad, Bea Alex, ...

[fedenanni]

@kmcdono2 don't worry and thanks a lot!