Experiments in Automated Entity Extraction with Pinpoint: Toledo Museum of Art Provenance PDF

Pinpoint is a tool for investigative journalists. It performs automatic entity extraction from PDF files. Can it be useful for processing provenance texts?

In this post, we examine the results for the Provenance Research PDF file published by the Toledo Museum of Art and archived at: 

https://web.archive.org/web/20121224083005if_/http://www.toledomuseum.org:80/wordpress/wp-content/uploads/Provenance-Research-lowres.pdf

Since the Toledo Museum of Art doesn't appear to publish provenance on its Website in 2024,  this older PDF file offers insights to the ownership history of artworks.

Using ChatGPT to detect language indicating uncertainty in provenance texts

In this conversation with the AI ChatBot, I asked it to look at the provenance text for Picasso's Woman with Pears, at the MoMa.

The provenance text on MoMa's website is:

[Galerie Kahnweiler, Paris and/or Ambroise Vollard, Paris and/or Galerie Thannhauser, Munich] 

By September 7, 1919 (1912 or 1913?) - at least October 1932, Alfred Flechtheim and/or Galerie Flechtheim, Düsseldorf, Berlin and London. 

By February 1934(?) - March 1937, The Mayor Gallery and/or Douglas Cooper, London. 

March 1937, Pierre Matisse Gallery (consignment no. 882, stock no. 594), New York, purchased from “Cooper”.  

March 31, 1937 - June 2, 1955, Walter P. Chrysler, Jr., New York, purchased from Pierre Matisse Gallery. 

June 2, 1955 - 1996, Florene May Marx (1903-1995, later Schoenborn) and Samuel A. Marx (1948–1964), Chicago, purchased from Walter P. Chrysler, Jr. through M. Knoedler & Co. (consignment no. 4786), New York. 

1996, The Museum of Modern Art, New York, acquired as bequest from Florene May Schoenborn.

I asked ChatGPT to break the provenance text into sentences and to identify any indications of uncertainty or speculation in each. Then I requested that it present the results in a table.

(I requested that it view "and/or" as uncertain.)

Sentence	Indications of Uncertainty or Speculation
[Galerie Kahnweiler, Paris and/or Ambroise Vollard, Paris and/or Galerie Thannhauser, Munich]	Indication of uncertainty in the gallery names
By September 7, 1919 (1912 or 1913?) - at least October 1932, Alfred Flechtheim and/or Galerie Flechtheim, Düsseldorf, Berlin, and London.	Uncertainty about the specific year (1912 or 1913) and presence of "and/or"
By February 1934(?) - March 1937, The Mayor Gallery and/or Douglas Cooper, London.	Uncertainty about the specific date (February 1934) and presence of "and/or"
March 1937, Pierre Matisse Gallery (consignment no. 882, stock no. 594), New York, purchased from "Cooper."	None
1996, The Museum of Modern Art, New York, acquired as bequest from Florene May Schoenborn.	None

Then I asked ChatGPT: 

what is the first year for which there is solid information

ChatGPT replied:

The first year for which there is solid information in the given text is 1937.