Generating oral history metadata digitally

Inspired by the amazing work that Dan Royles is doing with TEI mark up of oral history interview transcripts (soon to appear in Oral History Review) I took a quick run at something I first thought of at the workshop organized by the American Art History following the symposium, Digital Scholarship: New Avenues of Exploration in 2013. Like some other scholars, I trace my current digital history work back to my original practice of oral history.

The transcripts I wanted to work in 2013 with are now online, twenty interviews done for the Elizabeth Murray Oral History of Women in the Visual Arts Project by the Archives of American Art.

 I picked the art critic Lucy Lippard for this test run since I've already written about her involvement in early feminist art networks.  I analyzed her interview transcript by comparing it to the other transcripts in the project with a software called Antconc (a brief how to).  This computer program generates keywords, those words that appear more frequently in her interview transcript than in the others. I took the top 200 “key words” words and then cleaned them to create a list of people, places, and significant (to me) words. I then created a weighted, color-coded word cloud using Tagul which is fun, and is how JSTOR DFR displays similar data, but for metadata in a collection these could just as easily be presented as a linked list.

names, places, and key words generated by Antconc keyword analysis of Lippard's transcript

I was also curious about how well extracting entities might work for a more comprehensive sort of metadata, such as names, or a visualization, such as map of locations or organizations mentioned in all 20 transcripts.  I took a run at that as well, extracting names from Lippard's transcript using Stanford NER (more how to).

Names extracted from Lippard's transcript with Stanford  NER 

Already I see names familiar to me from an earlier essay, like Joyce Kozloff and Judy Chicago, and I'm intrigued by her world as represented by these names, those of prominent male critics like Lawrence Alloway and Hilton Kramer, and artists, Eva Hesse and May Stevens.

I also took locations extracted out of Lippard's transcript which I then dumped into Google map.  Carto DB would work much better but I needed to geo code it first.  As a nationally prominent, and internally known, art critic, I'm not too surprised by some of these locations, but once other transcripts were analyzed this might be a very interesting way to get beyond the NY/LA focus of so much feminist art movement scholarship.

Metadata needs will vary, but I could see pulling out the top 10 personal names, places, and key words for each transcript as a quick way to get an idea of the content. I can envision as well all sorts of ways to use this method to explore the discursive world of the twenty women in the Elizabeth Murray Oral History of Women in the Visual Arts Project.