Some things I have learned in recent days:

• data must be cleaned. Really. It’s probably still too noisy, even when you think it isn’t. Eliminate frequently occuring meta-notes (as it were). All citations to Guest & Wells on Coins in the UK, for instance, really muck things up.
• you can enter a single csv file as an input for MALLET. I knew this; but I had forgotten it, faced with a few hundred thousand rows of material (as I type this, the thought also occurs that I could run MALLET on the entire single file download I got from PAS, all ~500 000 rows. Presumably, locations and periods would sort themselves out into different topics?)

• MALLET considers letter characters to make up words. If you’ve got other stuff in there – numerals, for instance – that are significant, you’ll need to become familiar with – -token regex , which you’d use during that initial file-import. It was suggested to me to try these

— token-regex \s\d+\s

–token-regex ‘[\p{L}\p{M}\p{N}]+’

What else? Oh, that’s about all, for now. Oh, wait: custom stopwords. Instead of –remove-stopwords, you’ll want –extra-stopwords yourlist.txt . And your list has to be formatted so that there is whitespace between the words. I’m not sure if that means ‘white space’ like how you and I would figure it, or if that means ‘white space’ in some kind of crazy hidden code kind of way (like this in regex: \s (see this)). If you open one of the default stopword lists, there doesn’t look like there’s any hit-the-space-bar-kind-of white space that I’d normally assume.

Onwards!