Note: 12NOV2024 The Looted Art Detector is undergoing an update on Pythonanywhere and will be back in service soon
Objective: Identify high priority artworks for provenance research
Description: Online Free Digital Tool
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.
The resulting calculations do not signify that an artwork is looted. They simply quantify observations concerning the text for further analysis.
- The user uploads a CSV file that contains provenance texts
note: The uploaded CSV can contain other information as well - urls, titles, artists, etc. The only requirement is that it also contain one column with the provenance texts.
The program will ask the user to enter the name of the column that contains the provenance text.
- The Provenance Text Analyser calculates the number of times certain words appear in each provenance text and downloads a CSV named results.csv
note: The results.csv file contains all the original information uploaded by the user PLUS additional columns with the word counts.
- The user opens this enhanced CSV in his/her preferred spreadsheet or other tool and sorts and filters the results to obtain a priority list of artworks most likely to have problematic provenances
Instructions/suggestions are provided to help the user sort the results.
1) Recommendation: Create an "Uncertainty Index" and sort in descending order
Uncertainty index = Uncertainty Flags/Word Count
Explanation: The artworks with the highest concentration of words like "probably, "likely", "maybe", "possibly" and "?" show high uncertainty about the provenance.
- The user has the option of uploading his/her own custom list of key indicators (words and names to count)
Explanation: The list of key words determines what is counted and how it is categorised. This list is flexible and a work in progress. Users are encouraged to download the existing list and add their own words and names. Indicators of UNCERTAINTY, UNRELIABILITY and ANONYMITY change little from one area to another (in English), however Red Flag names will differ from Nazi-looted art to looted antiquities or forgeries.
FORMAT of the KEY INDICATORS (to count) file:
(content of default file at present)
The user is encouraged to copy this file, add words or names and upload it in the Looted Art Detector
Example of file that includes some Red Flag names from the Art Looting Investigation Unit Red Flag list as well as the names of some Jewish collectors who were plundered)
One notes that several new types of flags appear: flagALIUred, flag plundered, and flagRestitutionCase.
TO DOWNLOAD CSV below
The user is invited to copy or DOWNLOAD the above file, enhance it and use it in the Looted Art Detector in the optional step 3
. (3. (Optional) Replace default indicator file with custom CSV file)
About the Looted Art Detector tool
----
This free online Looted Art Detector tool uses code developed during the Swiss GLAMHACK2021 by Rae Knowler,
building on code developed during the Swiss GLAMHACK2020.
It is an experimental work-in-progress.
Please test and share your feedback and questions, which will be integrated into the FAQ
PythonAnywhere: https://artdata.pythonanywhere.com
Documentation
Github: https://github.com/parisdata/2021GLAMHACK
developer Glamhack2021: Rae Knowler
The Looted Art Detector online tool is based on code developed at the Swiss GLAMHACK2020 (team: Julia Beck, Birk Weiberg, Tamalera, Laurel Zuckerman)
Glamhack2020: https://www.openartdata.org/2020/06/art-provenance-dataset-text-analysis.html
*****
DATASETS with PROVENANCE to TEST
1. Detroit Institute of Arts European Paintings Acquired after 1932
DOWNLOAD CSV
No comments:
Post a Comment