Okay, I'm going to live-tweet another article! This time it's Goldstone, Andrew and Ted Underwood. "The Quiet Transformations of Literary Studies: What Thirteen Thousand Scholars Could Tell Us." New Literary History, vol. 45, no. 3, 2014, ideals.illinois.edu/bitstream/hand… #dhintro18
(Okay, to be fully truthful -- I'm not exactly live-tweeting it, as I read it several days ago, but everyone was tweeting about Kavanaugh and it felt rude to happily tweet about how neat this article is while that was happening)
Anyway, Goldstone and Underwood start out by arguing that literary study is often understood as being about conflicts (among ideas/institutions/generations), but this focus obscures other changes that have happened.
They also discuss the importance of context where language is concerned, and the potentially flattening effect of quantitative approaches. But their use of probabilistic topic modeling attempts to infer meaningful groupings.
Topic modeling isn't based on WORDS, it's based on TOPICS. This is SO COOL. A word like "nature" has different meanings in different contexts, so the algorithm takes that context into consideration by looking at the words near it. SO a word might show up in more than one topic!
Goldstone and Underwood spend a lot of time looking at different topics over time in seven journals. They use topic modeling to detect what they call the "quiet changes" that don't get theorized or noticed. They emphasize that this approach still requires interpretation!
Probabilistic topic modeling doesn't give the same results every time, and it's a *model.* That is, it's one way of understanding the data. It definitely doesn't provide a singular answer!
Topics aren't always easy to interpret! The algorithm puts together words that humans might not.
The authors find this approach "congenial to humanists" because of the interpretation required, and because multiple interpretations could be valid.
The authors find this approach "congenial to humanists" because of the interpretation required, and because multiple interpretations could be valid.
Goldstone and Underwood are cautious about attributing changes to specific writers and scholars. It's tempting to credit Foucault for all this talk about power, but he didn't start it!
They also write about how this kind of work can counter what they call "presentist bias," showing that certain ways of looking at literature emerged recently. They caution against rhetoric about returning to an imagined "core" of literary studies.
They have a lot of really interesting conclusions about how the talk about grammar has changed, the difference between scholarship and criticism, etc.
In any case, they emphasize that this is a methodological resource and definitely not a way to look at literature "objectively."
AND they share their data! Here are their topics: rci.rutgers.edu/~ag978/quiet/ This is my favorite part and I want to spend some more time playing with it.
Unfortunately, I had some trouble getting Mallet to work and so didn't end up using it for my text mining experiment -- but I think I'll keep trying. It might be a better approach for what I wrote about in any case. #dhintro18
