Engl507

Log 4: Topic Modelling a Text

For this entry, I'm asking you to run this MALLET topic modelling tool at least five times, using different parameters, through a text or collection of texts of your choice. Based on my experience, I recommend using TXT files. If you prefer, then you are more than welcome (encouraged, even) to operate MALLET via the command line.

In your log entry, please:

• Push the results of all five trials to your GitHub repo. For each iteration, please include (at least) Topics_Words.csv and TopicsinDocs.csv. I don't need to see the HTML output, but you're of course welcome to archive it.
• In writing, compare your results, drawing examples from the topics. What congeals across each? What stands out? What's curious? Where are the slippages and carrot sticks?
• Finally, give me your take on topic modelling and its relation to surface and/or distant reading. How (if at all) would you use topic modelling to articulate an argument? How much do you (or should you) trust MALLET? What surprises does, or could, it foster? What are its limits, assumptions, and blindspots?

No need to write much here. Between 250 and 350 words is perfectly fine. Also, as you write, you might want to consult this page, which contains an overview of MALLET and some related resources, including other scholars' perspectives on topic modelling.

Let me know what questions or concerns you have!