Textual material is the foundation of humanistic inquiry and discussion. Essays, scripts, and novels are useful tools for communicating ideas and arguments. Thus, the field of Digital Humanities must formulate approaches for computationally analyzing textual data and the networked relationships within them. One way to get a feel for the range of approaches available for textual analysis is to explore visualization tools in programs like Voyant. My investigation of Voyant Tools and its JavaScript-driven application changed my mind about how I believe the field should progress and what tools deserve the largest emphasis.
My investigation began with experimenting on the Voyant Tools web app. I chose to analyze the Wizard of Oz script because I’m familiar with the characters and storyline. The tool that immediately caught my eye was the word cloud or “Cirrus”. Largest on the cloud were terms like “Dorothy” and “witch”. Interestingly, terms like “shall” and “asked” were also present. These clouds jumble important terms with words that have no meaning without their context. Now, I tried to locate tools on Voyant that I felt better represented the context and information contained in the script. The TermsBerry tool lists high frequency terms and words that are often close in proximity to the original terms. It displays this data in a handy visualization tool that allows you to hover over a word to view its collocates and relative frequency. Tools like TermsBerry provide critical context to terms like “shall” and “asked” and give the basic word cloud visualization tool extra depth.
Next, I moved to explore Voyant Tools through the Spyre Notebook JavaScript tool. This tool allows access to Voyant Tools through a JavaScript interface. My exploration of Voyant through JavaScript led me to the Bubbles tool. This tool tracks the script chronologically and increases the size of frequent terms as they are read. I felt that the Bubbles tool did well to illustrate the development of a story like the Wizard of Oz but lacked some of the context that TermsBerry portrayed.
After analyzing Voyant Tools on the web and through JavaScript, I believe that we should be wary of textual analysis that strips the content of its context and progression. AI summaries and text analysis can be misleading. They can describe the wrong argument or not tell the complete picture of a text. When considering the wide range of analysis approaches that exist, I suggest that digital humanists aim to focus their attention on tools like TermsBerry and Bubbles that illustrate context and continuity.
I wrote something similar about the AI being misleading. I find it a little counterintuitive that one must really know the texts for this too work that well. At that point, we might as well make close-reading interpretations as humans rather than distant reading with AI. I found that Voyant wasn’t giving me much information about Shakespeare’s the tempest, but I think that the Bubble tool giving chronological information is the closest that a tool can get to breaking down the story somewhat. I can’t help but wonder how the confusing the Bubbles would look like for a nonlinear plot.