Topic modeling, and the greater umbrella of text analysis, is being used more frequently in humanities scholarship. Using a computer to take a distant look at many texts in order to see patterns, dive deeper into the texts, and interpret these results provide an ever-changing framework for humanities discourse. However, at what point must we consciously focus on keeping the “humanities” in “digital humanities?”
Throughout the previous year, I went straight for that gap between traditional and digital scholarship. I studied poetic content in The Star of the North; a Pennsylvania newspaper printed during the U.S. Civil War, available through Chronicling America, the Library of Congress’ initiative to digitize historical United States newspapers. Specifically, I utilized the topic modeling software MALLET to compare themes of poetic verse to those of the remaining textual content, often news articles and prose. I created the algorithms, ran 298 poems and 483 pages through them, and watched as the computer returned sets of words. At that point, it was my task, as the humanities scholar, to interpret those sets, and form my own conclusions about the topics they created, and what that said about these newspaper issues during the U.S. Civil War.
While I could make a persuasive case about the textual content of The Star of the North from my results and close reading, which would interest humanities scholars, such as poets, U.S. Civil War literature scholars, and newspaper scholars, among others, I could not have arrived at this knowledge without the distant pattern reading that topic modeling allows. The digital process pointed out topics I wouldn’t have expect to appear, as well as interesting anomalies, which provide an even better view of the text. Topic modeling didn’t give me any answers; it gave me the tools to seek out the answers myself in an innovative way.
This realization, along with my awareness as a human moving through this digital humanities case study, proved to me that, even through the frenzy of everything going digital, there is still incomparable importance placed on traditional scholarship. I arrived at the conclusion that digital scholarship should not replace traditional scholarship. Rather, digital scholarship can provide a space for the traditional scholar to imagine even more, further conclusions than traditional close reading can offer.
MALLET documentation: McCallum, Andrew Kachites. "MALLET: A Machine Learning for Language Toolkit." http://mallet.cs.umass.edu. 2002.