Comparing Tools | Voyant, Kepler, & Palladio

In a nutshell, what do these tools do?

• • Voyant is a web-based tool that reads large bodies of text and visualizes the most common words via computer-generated text analyses.
    Visualizations include: Cirrus, Reader  (digitized text excerpts), Trends (in the form of a graph),
  • Kepler is an open-source geospatial analysis tool takes a collection of data and populates it on a map so that you’re able to visualize connections, trace trends, and analyze more deeply (in a possibly new direction, too).
    Maps include: Point, Arc, Line, Grid, Hexbin, Polygon, Cluster, Icon, Heatmap, H3, 3D, Trip
  • Palladio is an easy-to-use web-based visualization tool that populates data on a map, in a network graph, in a table, and/or in a gallery.
    Visualizations include: Graphs, Maps, Lists, and Galleries

Does comparing what the tools reveal about the narratives highlight anything else about what you can discover using them?

I mentioned in one of my posts that I was a little embarrassed that I made so many modern assumptions on a historical artifact like the narratives. I thought, How amazing that we have these narratives as part of our national archive! I’m sure that the interviewers diversified their subjects, making sure to talk to all sorts of folks, from all sorts of backgrounds to ensure that they were getting a comprehensive snapshot of this story.

Of course, that’s not at all true, which both Sharon Ann Musher’s piece highlighted and Voyant, Kepler, and Palladio started to validate. These interviews were conducted by primarily white interviewers in the Jim Crow South where racial tensions were still insanely high and former slaves were asked to recall an undoubtedly emotional memory without any trepidation or filters.

This partiality is a theme that I touched on a good bit in the last three modules, mainly because it stood out as I worked through the visualizations, but it also got me thinking about what this dataset would look like if you approached it as a known bias and used Voyant, Kepler, and Palladio to unpack the data elements in such a way that you try to find the other side of the story or at least try to find the breakdowns in interview diversity. Is it the state that had the biases? The interviewer? The interview subject? Did where the former slave was enslaved skew their recollections?

Possible exploration paths to uncover this side of the narratives could be to:

• • Analyze the popular (if not cliché) words using Voyant
    • Select a dozen high-profile words in Voyant that you’d expect to see when reading former slave interviews. Track their popularity across each state.
    • Start to look at each state to track high frequency words.
    • Analyze each state’s top words (3-6, let’s say) and see if anything stands out – are there any unexpected words that crop up in one state over another? Is there a word that seems fabricated or exaggerated, especially when seen in context using Voyant’s Reader tool?
  • Analyze the interviewers using Palladio
    • Dig into individual interview patterns
      • Is there an obvious partiality with the way each interviewer conducted their interviews? Did they stay in one location? Did they favor urban over rural or rural over urban?
      • Did certain interviewers only interview certain types of former slaves? Did they even ask the question, or did they gloss over that side of slavery and just categorize the interview as unknown?
  • Analyze the former slave’s journey from where they were enslaved to where their interview was conducted using Kepler
    • Trace the location of enslavement to where the interview was conducted
      • Is there a theme between certain locations of enslavement and where these slaves moved once they were freed?
    • Analyze a city’s interviews for trends
      • Is there a common location of enslavement for interview subjects living in a certain city? Did a lot of slaves from Virginia move to Mobile, Alabama, for example?

Are there ways that tools can complement each other?

The question, however, is not how to write the history of information but what to do with all of it, as historians in an era when we have a burgeoning array of techniques and technologies available to help us scrape, mine, curate, analyze, and visualize data.
– Historical Research in a Digital Age: Reflections from the Mapping the Republic of Letters Project

The short answer is yes, these tools can very much complement one another. For example, let’s say that I took some of the questions from above and started to explore this dataset using a more narrow scope in Kepler, Palladio, and Voyant. I start with a basic text analysis in Voyant, compare this analysis to the network of potential biases on the part of the interviewer in Palladio, and then ultimately juxtapose both of these analyses to where former slaves were enslaved and where they were interviewed using Kepler. I suspect that these analyses could provide a lot of insight into somewhat hidden meanings in the original narrative transcriptions.

While each tool individually is interesting and thought-provoking, imagine what story could surface if you divided the data to conquer this idea of partiality in the interviews and how location (of enslavement and interviews) told a very lopsided story of a very difficult time in American history.