For this project, I chose to look at data mined from a small set of Old Norse sagas (both modern retellings and medieval text translations) and from a set of 30 texts related to Robin Hood, most of which were fictional retellings of the early Robin Hood ballads.
I used Project Gutenberg as the digital source for the Old Norse sagas and cleaned up the texts slightly by deleting some irrelevant introductory material before running them through Voyant. I had read Stanley Fish’s terse “Mind Your P’s and B’s” column before loading anything into Voyant, and, determined to hypothesize prior to experimentation, I wrote several test questions. Here they are, in no particular order:
- What can Voyant tell me about the role of music in Old Norse sagas?
- In stories of Norse heroism, are kings or gods considered more important to the plot, and is that import consistently portrayed in both medieval texts and modern versions of the sagas?
- Will Voyant’s data show evidence of the practice of boasting through an abnormally high number of personal pronouns associated with feasting?
- How often are boats and seafaring mentioned in these works, and what can travel-related words tell me about Norse ideas of honor and independence?
Unfortunately, the only question I could begin to answer with Voyant’s data is the last one. After analyzing all the sources, I realized that I had been asking the wrong type of questions, as my questions couldn’t be answered satisfactorily with relative frequency data. Or, at least, I needed to start with a more basic question and see if I could trace the thread of a syntactical pattern or unique contextual use first, and then follow that thread to a deeper conclusion. This type of dependency on data—a sort of guided treasure hunt instead of a quest to prove or disprove a hypothesis—is, I suspect, why Fish seems to have a problem with data-based literary studies. After completing my analysis, though, I did learn something about Old Norse sagas, even though it wasn’t what I expected to learn, and I count that discovery as something just as valid as a proven hypothesis: if read through their most commonly used action verbs, the Norse sagas are stories of exploratory movement and speech, more focused on ship-speak and interpersonal communication than the violent hand-to-hand combat that seems to drive each protagonist’s storyline. For example, in Eirik the Red, the most common action verbs are came, went, said, called, set, took, sailed, received, prepared, and gave. In the English version of Snorri’s Heimskringla, the most common action verbs are came, went, said, took, called, come, told, gave, says, sent. This distant reading actually spawned close reading, as I started to scrutinize the texts to see if a microanalysis of the stories evidenced a similar conclusion. In summary, it did—the dramatic climaxes of the sagas paled a bit when I began noticing the emphasis on a give-and-take speech peppered throughout the works. It prompted me to think of Norse women and men as members of communities who prized societal justice and common welfare as well as individual bravery.
Pleased with the results from Voyant, I chose to run the Meandre_Topic_Modeling algorithm in HathiTrust in order to examine my Robin Hood sources. This time, I ran the algorithm before posing a research question (somehow, I could feel Fish’s disapproval), just to see what patterns emerged. I wasn’t surprised at the topic groupings, though: the word men was weighted heavily in more than one cluster (due to the repeated use of the phrase merry men, I wager); the words archers and arrows were closely associated; the word lady appeared in conjunction with true, days, fair, and sweet, etc. It occurred to me that this particular algorithm might be more useful when trying to discover topics from a larger set of slightly more diverse works. For instance, if I had included more literary criticism instead of just loading several copies of the same ballads, I might have been able to see interesting groupings of critical themes that could be used to plot the wax and wane of scholarly interests in relation to medieval outlaws.
I ran the Simple_Deployable_Word_Count algorithm using the RH texts, as well, and was rewarded with a short list of the most common words in the early ballads. As expected, most of the words were articles and prepositions; nevertheless, one surprising term appeared, namely, quotation marks (counted as a word in this algorithm). The quotation marks highlight the colloquial nature of the ballads and emphasize their vernacular media—as dialogue in tavern songs and popular rhymes—but also speak to the value of the written text. (What need does a minstrel have of quotation marks?) The ballads were preserved through ink, and the added punctuation is a testament to the endearing legacy of the ballads and the complicated shift towards a more literate populace near the dawn of the Renaissance. It would be interesting to compare punctuation in medieval and Renaissance versions of the ballads with an algorithm such as this.
