Kyle Rawlins

linguistics & music, not necessarily in that order. Assoc. professor at @JHUCogSci; professional website:

[Kyle Rawlins @icoson]

one more try, now relocated to https://t.co/MMgJikcDsV

[Kyle Rawlins @icoson]

@HadasKotek @thai101 @GretchenAMcC @lingstitute2023 @anggarrgoon Alright; I mention it because given that discussion there it’s a bit unclear if the organizers are ok with there being an unofficial one, but it looks like @anggarrgoon is on that question

[Kyle Rawlins @icoson]

@thai101 @GretchenAMcC @HadasKotek @lingstitute2023 @anggarrgoon There’s a discussion about this in a thread of the prof slack in the general channel (it’s all in response to a Q about Facebook, so may be easy to miss)

[Kyle Rawlins @icoson]

@patrickdelliott contextual domain restriction is powerful

[Elias Stengel-Eskin @EliasEskin]

Language is ambiguous, but NLP often makes the simplifying assumption that it's not.

In work with @icoson and @ben_vandurme we introduce ✨AmP✨, a new framework for studying ambiguity in semantic parsing, esp. with LLMs

https://t.co/All5DL3SWn

(arxiv link to come)

1/10

[Kyle Rawlins @icoson]

@HadasKotek I actually think there’s already plenty of work out there that is quite interesting but notably not stuff that confuses models and theories — per a classic 1st wave connectionism paper few people now seem to have read/remembered, https://t.co/PfgEmCbAop

[Kyle Rawlins @icoson]

one more try, now relocated to https://t.co/MMgJikcDsV

[Kyle Rawlins @icoson]

[Updated] Like a lot of people this days, diversifying my social media a bit: trying out mastodon again (currently @[email protected]), also cohost (icoson)

[Linguistic Society of America @LingSocAm]

⏰ Fellowships are now open for this year's Linguistic Institute, taking place June 19-July 14 at @UMassAmherst!

This year's theme is “Linguistics as Cognitive Science: Universality and Variation.”

More info: https://t.co/oe2ZmYKHZc

@lingstitute2023

[lingstitute2023 @lingstitute2023]

We are happy to announce that we just published the courses in the program of the 2023 LSA Summer Institute! Thank you to everyone who submitted proposals and helped in the review process. We received over 170 applications! https://t.co/g51HFhNeLq

[JHU Cognitive Science @JHUCogSci]

We are working on setting up an alternative account as JHUCogSci at sigmoid dot social (now to see if this tweet will go through)

[Kyle Rawlins @icoson]

@bronwyn I’ve known lots of successful cases (incl me, and family members in a different household), so it’s definitely worth trying; hard to know what exactly causally helped but I wore n95 masks consistently in shared spaces

[Kyle Rawlins @icoson]

@LedermanHarvey I just did some googling based on this tweet and the top hits were skepticism about dessert hummus which ... seems pretty reasonable actually

[Kyle Rawlins @icoson]

@kmahowald @DruckerDaniel yeah, one conclusion I came to while thinking about this more is that the "in X respect(s)/way(s)" construction requires sufficient mereology of some kind (multifacetedness?) in its argument to be licensed, and for beliefs this may or may not be provided by the thing believed

[Kyle Rawlins @icoson]

@DruckerDaniel FWIW I would accept "I believed his claim in one respect but not another" (at least, probably given sufficient context)

[Kyle Rawlins @icoson]

@DruckerDaniel more concretely, you certainly can believe various things in some respects or in one respect but no another. It's just that propositions aren't one of those things. So maybe it is a claim about propositions after all: propositions aren't comparable to people (in NL at least).

[Kyle Rawlins @icoson]

@DruckerDaniel Not sure I understand the question, sorry! But this isn't intended as a claim about propositions, but more about the (non-specialist) selectional patterns of "believe" and "in some respects".

[Kyle Rawlins @icoson]

@DruckerDaniel I haven’t read all of this thread, sorry if redundant: in natural language, we don’t believe propositions (as nominal arguments); we believe rumors, stories, claims, headlines, […]. This is what is comparable to “admiring a person”. “Believe the story in some respects” is fine

[weird medieval guys @WeirdMedieval]

oh no

[Kyle Rawlins @icoson]

@rhenderson @ToddTheLinguist sorry; seeing other replies and dbl checking, this isn't IFA, I was looking too quickly. This is more analogous to e.g. the world-aware FA rule in (16) of the vF&H thing I linked. (IFA abstracts over a world parameter and then does FA.)

[Kyle Rawlins @icoson]

@rhenderson @ToddTheLinguist The more normal version of IFA (at the very end of H&K, ex (25) of https://t.co/wEgqBHErxz) has worlds as evaluation parameters, but this is recognizably very similar

[Kyle Rawlins @icoson]

@brianbuccola @HadasKotek @themitcho yeah, that's fair; I do have a specific conda env that I use only for jekyll and otherwise leave exactly as is