In the past few months there has been a lot of attention paid to a Zotero plugin called Paper Machines. Paper machines was coded by Brown grad student Chris Johnson-Roberson, and Jo Guldi and Matthew Battles directed the project. Paper Machines uses the data in a Zotero collection to generate analyses and visualizations. If you have a sizeable collection of documents with good metadata and full text (for example, PDFs with text layers) then Paper Machines can run its analysis. The most basic output is a word cloud. More advanced analyses provide phrases matching a pattern, maps of place names, annotations of people, places, and organizations. Paper Machines can also perform topic modeling using MALLET.
I ran Paper Machines on the digitized primary sources for a chapter of my dissertation. I didn’t do any work to clean up the data, and this is obviously just a first pass without any analysis. But I was impressed by this map of place names, which generated the kind of American-British-German connection that I thought I saw in the sources. But the plot also contains places that I didn’t expect, which I hope will be a fruitful area to research further.
For a much fuller description of Paper Machines, you can see the documentation, Johnson-Roberson’s write up at Harvard’s metaLAB, or a two blog posts by Sarita Alami from the Emory libraries.
Have you tried Paper Machines with your Zotero library?
Update (11/8): In the original post I neglected to give credit to Jo Guldi and Matthew Battles who started and directed the project. You can read a post about Paper Machines by Guldi here.
Update (11/9): Chris Johnson-Roberson sends this update about the contributions of the various collaborators: “Paper Machines was coded by Brown grad student Chris Johnson-Roberson in collaboration with historian Jo Guldi, under the guidance of Matthew Battles at Harvard’s metaLAB.”
