"There might be a good reason why no-one has filled the gap – it’s not interesting, it’s not significant. Not being done is not enough of a reason to do a piece of research." blogs.lse.ac.uk/impactofsocial…

Just in time for your late summer writing frenzy – the preliminary call for the Society for Computation in Linguistics, deadline Oct. 4, conference held virtually Feb. 7-11. Allyson Ettinger @bpricket blogs.umass.edu/scil/scil-2022…

If you thought this sounded interesting... surprise, I think we might have one! timhunter.humspace.ucla.edu/papers/FrankHu…

In medieval Europe, 90% of the population needed to be farmers. Today, we can accomplish the same trick with just 2% of humanity standing around in the fields. That means that we now have the labor to staff the healthcare system. humanprogress.org/the-secret-rec…

It almost always seems to pay off to remind ourselves, one more time, that assigning probability mass to items not seen during training just *is* departing from complete alignment with the training data.

We know that for any NFA with n states, there is an equivalent DFA with at most 2^n states. Question: if I have a DFA with 2^n states, under what conditions is there an equivalent NFA with exactly n states? I've been poking around but cannot seem to find anything about this.

This is what Dana Angluin would tell her students at the beginning of her discrete math class at Yale, which I TAed back in grad school. It was always quite a shocking thing for them to hear when they had just been taught the opposite in high school!

In Africa, "healthy life expectancy—or the number of years an individual is in a good state of health—increased to 56 years in 2019, compared with 46 in 2000." reliefweb.int/report/world/h…

Over the last few decades, life expectancy has risen across all country income levels. Importantly, the gap between the richest and poorest places is narrowing. Explore the data: humanprogress.org/dataset/life-e…

Given a weighted CFG, is there a way to sample from the globally-normalized distribution over derivations (where a tree's prob is proportional to its score), *without* first converting to the locally-normalized version of the grammar?

Family With 2-Hour Layover Sets Up Rough Shantytown At Airport Gate bit.ly/3cx5bH5

Please see our RFP, we are looking for a search firm to hire a new executive vice-provost for vice-provost recruitment.

Norbert in fine form: "Chomsky’s point is that sometimes nothing actually is better than something, since postulated somethings that don’t do any explanatory work actually blind us to our ignorance"

"Cynicism has become a hollow shorthand for intelligence. Afraid of being labeled naive, too often our discourse is characterized by hostile, knee-jerk skepticism that doesn’t allow much room for doubt or wonder." - Brandon Stewart for The Progress Issue

Jobs: Assistant or Associate Professor of Syntax, University of California, Los Angeles: The UCLA Department of Linguistics seeks to fill a faculty position (tenured or tenure-track), starting July 1, 2023, for a specialist in contemporary formal… dlvr.it/SZ5pn1

Probabilistic Exact Recursive Programming Language youtu.be/NkeCo4UfYAI

HelloFresh Announces Collaboration To Discreetly Deliver McDonald’s In Its Packaging bit.ly/3VaKFgT

The program for the Society for Computation in Linguistics (June 15-17) is available: blogs.umass.edu/scil/scil-2023… Lots of cool-looking stuff, thanks Joe Pater, Tim Hunter and @bpricket for organizing!

If you're going to UMass for the LSA summer institute, consider coming a few days early for the SCiL conference (June 15-17): blogs.umass.edu/scil/scil-2023… lingstitute2023 #LSAInstitute2023