Looted Art Detector


The Looted Art Detector is a project in progress. Feedback welcome. 

Try it on any CSV file!


For instructions and tips on using this digital tool, please see:

Looted Art Detector Tool: Swiss GLAMHACK2021


Approach: Automatic text analysis using frequency counts
note: The frequency counts target textual indicators of UNCERTAINTY, UNRELIABILITY,  or ANONYMITY, as well as the possible presence of RED FLAG names related to NAZI-looted art, forced sales and duress sales. You can add your own custom lists of words or phrases to count.
The resulting calculations do not signify that an artwork is looted. They simply quantify observations concerning the text for further analysis. 



Uncertainty and the Looted Art Detector
In default mode, the Looted Art Detector counts indicators of Uncertainty, Unreliability or Anonymity. Why uncertainty? 
Because words like "probably", "likely", "possibly", "maybe" may signal that statements in the provenance text are uncertain, speculative or unsupported by the evidence. A high concentration of such words may indicate a text that is deceptive or false. 

One of the reasons that false provenances are published, in particular for NEPIP items, is to conceal an ownership history that is linked to the Holocaust. 


Tutorial for the Looted Art Detector: Using custom indicators


The user can analyse provenances for any names or words that seem interesting. You can upload your own lists of words to verify. It's easy. This post shows how to do it.



Detector Tool Tutorial: ranking by number of Red Flag and Restitution Case Names


The ranking - which will of course change depending on which criteria one chooses to use -  should be both transparent and replicable by others. This post looks at how to use the output file to rank artworks by the probability that they have a problematic provenance.


page in progress 

No comments: