Zekka

but sorry, i totally made the convo about me while trying to explain why 90% of my statements about language are going to be terminologically garbled or misinformed in some way!

it is, but i'm increasingly trying to turn us towards using some already-built machine learning models on the basis that I will probably not be able to tune the system we end up using

(when i say "my background in language" i mean "help my boss is making me do statistics on natural language data even though that's not technically in my job description")

see you, good luck!

oh yeah, i've used both models. i'm kind of a markov chain fan just because they're easy to implement and don't require a ton of mathematics, but tbh sparsity is still a real problem for me

(basically, giddy addled zekka thought "oh! they mean likelihood according to a markov chain, not according to a bag of words" where they actualy meant neither)

but I think "is allowed to occur" and "is likely" are similar enough that if you sub "is allowed to occur" for "is probable" in my sentences some of them would still be true

oh. sorry, my background in language is all statistical so when I saw a familiar word I latched onto it

if you say "well, a coordinated pronoun like A and B is only OK if A would be OK or B would be OK" then you also mispredict that. and you don't account for how "B and A" is often more likely than "A and B" (x: he and I vs I and he)

your argument is that if you say he/him-ness is a completely independent characteristic from he/you/I-ness, then you expect there to be as many "whom"s to "who"s as there are "him"s to "he"s

oh hey thanks for writing more details! so by distribution you basically mean "odds of happening based on other features of the sentence"

rephrasing to make sure i got it: some people assume that if "me" happens half as much as "I", then "him" probably happens half as much as "he." sorry, i struggled to rephrase the coordination thoughts but i'm not very confident i understand what it means

hey, i think the case distribution observation is really interesting! i don't think I've ever written down that assumption, but I definitely think I've made it implicitly.