In an op-ed today for MIT Technology Review, Elemental Cognition's Jesse Dunietz traces recent soul-searching in #NLProc to questionable evaluation methodologies, and describes our new approach to assessing machines' reading comprehension. technologyreview.com/2020/07/31/100…

Before PCA (i.e SVD), I preprocess with three principals: 1) sqrt any features that are counts. log any feature with a heavy tail. 2) localization is noise. *regularize* when you normalize. 3) and my favorite rule, the Cheshire cat rule explanations in 🧵...

Asking about a fancy stats analysis for a manuscript vm.tiktok.com/ZMJk3Kwep/ #AcademicTwitter #AcademicTikTok

I can’t say enough good things about this free NLP bootcamp by CMU School of Computer Science Learn FOR FREE how to accomplish some impressive NLP tasks to benefit low-resource languages. github.com/neubig/lowreso…

Is it just I or anyone else having trouble adjusting to changes in schedules? It took me some time to figure out a daily schedule in May and then with classes starting I need to figure out this again? How did you adapt to this? Academic Chatter™ #AcademicTwitter

I just spent my entire afternoon figuring out why my bash script isn't working on a remote machine. Turns out I had wrong IDE setting on my local

Some talent cannot be encountered again. His songs were lively, relatable and inspiring. What set him apart was his love for languages and educating others about it. He lives on through his voice. #RIPSPB

Upcoming webinar cycle on The Science(s) of the Face -- organized by me on behalf of the ERC project FACETS. Attendance is free, everybody is welcome!

Larkin and Simon's "Why a Diagram is (Sometimes) Worth Ten Thousand Words" and the curse of dimensionality could give a better understanding of degrees of freedom in high dimensional problems?!

I use the story of the Boy Who Cried Wolf story to distinguish between type I and type II errors. The first few times, when the villagers come running even though there is no wolf, is a type I error; the last time, when they don't come but there is a wolf, is a type II error.

is this going to be like an F1 pit crew action?

See our work on predicting facial expressions in infants. Opensource code coming soon on github.com/AffectAnalysis…!

A Comprehensive Review of Data-Driven Co-Speech Gesture Generation Simbarashe Nyatsanga,🇺🇦 Taras Kucherenko @[email protected] , Chaitanya Ahuja, Gustav Eje Henter, Michael Neff tl;dr: if you are starting to work in gesture recognition/generation, that is a good read to start. arxiv.org/abs/2301.05339

I just solved these puzzles from Sasha Rush and as a person with no background in Cuda and GPU programming, it was very clear and fun. I LOVED it! I wish there were more learning resources like this available. github.com/srush/GPU-Puzz…

How to write a rebuttal for a conference? Writing an effective rebuttal helps answer questions, address reviewers' concerns, clarify misunderstandings, and help the AC make an informed decision. But it takes work to write a good one. 😟 Sharing some tips I found useful. 🧵

Our lab computers were down today, and I happen to have come across Mark Rober's commencement speech at MIT. I think his 3 point advice deserves a wider audience. I certainly found it useful. I hope you do too. youtube.com/watch?v=0FGlsu…

Finally, someone did it. Python for React. React is the most popular front-end framework used to build interfaces and now all python devs can use it. This means you can code an ML model, develop a backend and design a front end all in one language. github.com/reactive-pytho…

Reminder, AMAR workshop deadline is in one month on March 18th. cse.usf.edu/~tjneal/AMAR20…

AMAR workshop deadline has been extended until March 28th with decisions sent out on April 12th. We look forward to your submissions!