When I started in #ComputationalLinguistics / #NLProc, some of the stuff with word embeddings and topic modeling felt like kind of a stretch re: what we could do with language models, but at least the formalisms were interesting and it felt like there was an effort to treat it as exploratory and not "hey language models understand language!" Now it feels like the field has been co-opted by all the #MathyMaths and #TESCREAL bullshit, and I so desperately want to get back to real #CompLing