The Wind is a-blowing

The readings for this week focused on the question of textual analysis, for which I was grateful, given the current direction of my project. As we are getting deeper into the semester, my plans are growing more and more concrete, and I expect to have a large project update next week, to fully flesh out my intentions.
As such, the focus on textual analysis was of particular benefit to me this week, confirming many of my preconceptions regarding its utility and its limitations. Each of the four articles assigned shared a singular caveat, however. Textual analysis, though powerful and undoubtedly useful, remains a supplementary tool to traditional research methods. This is entirely unsurprising, given the level of training and historical immersion required to accurately assess period-specific documents for a professional historian. Judgment calls are required on a regular basis, and a computer or algorithm, no matter how sophisticated, remains limited in its ability to assess satire, emotion, and ambivalence.
With these limitations in mind, I found myself interested in the potential of topic modeling, a combination of textual analysis and visualization methods. Two of the assigned articles focused on very different approaches to this topic. The first was “Mining the Dispatch,” an assessment of the issues of the Richmond Dispatch during the American Civil War. The intent was to recreate a certain sense of everyday life from the accumulated data of an extremely large corpus. The limitations of topic modeling, of course, derive from not only the inherent issues with textual analysis, but the historian’s choice of the topics themselves. This project provides some illuminating data points (the dramatic uptick in the mention of war bonds towards the end of the war was particularly striking), but if it were connected to a more qualitative analysis (say, of the aesthetic qualities of the newspapers themselves), it would be even more effective.
The other large example provided was of the correspondence of Henry Kissinger, which was presented in a very different means of visualization. Here, textual analysis sought to illustrate the relationships between words (“bombing” and “Vietnam,” for example), and cluster them together. This had the benefit of not only being interesting, but visually accessible and appealing. Many visualization tools we have engaged with in class do not lend themselves towards legibility, and it can be difficult to recognize patterns even when they do exist (which, of course, is not always guaranteed).
Given my interest in textual analysis, Megan Brett’s Topic Modeling: A Basic Introduction was a good shorthand into the requirements of such a project: A large corpus which has been “tokenized” for computer legibility (i.e. the removal of capital letters, articles, etc); a familiarity with the corpus on the part of the researcher; a program or tool for the topic modeling; and a means of visualization which will make the research accessible upon completion.
My own dissertation, I am convinced, will benefit from some form of textual analysis, especially given the sheer amount of documents I have to work with. While this method remains a supplementary one, it can undoubtedly demonstrate and reveal previously unnoticed patterns and relationships.

Leave a Reply

Your email address will not be published. Required fields are marked *