marty_with_an_e's profile picture. Assistant prof of linguistics at @Cornell. I study humans and neural networks to learn about incremental lang processing strategies.

Accepting linguistics PhDs

Marten van Schijndel

@marty_with_an_e

Assistant prof of linguistics at @Cornell. I study humans and neural networks to learn about incremental lang processing strategies. Accepting linguistics PhDs

Pinned

Do you want RNN surprisal values for your data? I've got code to get them, which now comes with links to pre-trained models! (email me if you'd like your model added to the list) Thanks @bruno_nicenboim and @ArmeniKristijan for the prod github.com/vansky/neural-…


Marten van Schijndel reposted

UW-Madison Psychology is hiring! Two new assistant professor positions in *any* area of social or developmental science! Come join us!!! jobs.hr.wisc.edu/en-us/job/5105…


Marten van Schijndel reposted

✨ I'm on the job market! ✨ I translate methods from natural language processing to applications in computational social science. I measure model instabilities on datasets + study how people write about personal experiences (giving birth, reading books) in online communities.


Marten van Schijndel reposted

If you're applying to grad school this cycle, here's a plug for BU Linguistics (@BULinguistics)! I plan to participate in recruitment this year and would love to see applications from students who share my interests at the intersection of Ling/CogSci and NLP. _Very_ broadly,


Will Timkey (@wtimkey8) and I have a new paper at #EMNLP2021 showing that cosine similarity doesn't work for transformers. Luckily, simple post-processing techniques fix the issue. See quoted 🧵 for details.

Have you or a loved one used similarity measures like cosine similarity or L2 distance in transformer LMs? Our #EMNLP2021 paper shows that a few "rogue" dimensions consistently break sim. metrics in these models. Luckily, there are some easy fixes (🧵) arxiv.org/abs/2109.04404

wtimkey8's tweet image. Have you or a loved one used similarity measures like cosine similarity or L2 distance in transformer LMs?

Our #EMNLP2021 paper shows that a few "rogue" dimensions consistently break sim. metrics in these models.
Luckily, there are some easy fixes (🧵)
arxiv.org/abs/2109.04404


Marten van Schijndel reposted

We want you! Shoot me a DM or email me, [email protected]!

Are you early career (student, postdoc, pre-ten fac, early research scientist) working on #discourse in #aphasia? We'd *love* to host a talk from you through @FOQUSAphasia this fall! Please be in touch on here, or via email, [email protected]. Working on obtaining honorariums!



Marten van Schijndel reposted

I'm really excited to say that we are engaging in our largest hiring effort in the history of our department. Come join CSE @ OSU! cse.osu.edu/faculty-recrui…


Marten van Schijndel reposted

Hey there PhD students in cogsci, linguistics, or related field! Do you study how people learn? Are you graduating in 2021-22? Want a job where you can apply your knowledge to building learning tools used by millions of learners? Apply to work @duolingo! boards.greenhouse.io/duolingo/jobs/…


This is such important work. If you are interested in modeling human language processing, you need to check out the work @coryshain is doing. His CDR(NN) methods provide a way out of many of the simplifying modeling assumptions psycholinguists make for computational convenience.

Here's the pre-recorded talk on a deep neural regression technique for human subjects data, happening in < 2hr (16:30UTC/12:30EST) #ACL2021NLP session 11B. underline.io/events/167/ses… Questions welcome! (you have to be registered 😡; will post a public link after the conference!)



Marten van Schijndel reposted

Considering a submission to Glossa Psycholinguistics? We are a Fair Open Access journal, and we’ve adjusted our procedures so authors are asked to disclose their ability to pay only if their submission is accepted for publication.


Marten van Schijndel reposted

Interested in doing research on neural NLG? Still considering candidates for a post-doctoral scholar post on a Facebook-sponsored project at OSU. Talk to me at #ACL2021NLP! @osunlp u.osu.edu/white.1240/202…


Happening now at poster 2889 on gather.town. #ACL2021NLP

New paper w/ @Forrest_L_Davis! Do neural language models know too much language? We know: Shallow non-ling heuristics can overwhelm linguistics in LMs We show: Ling knowledge also overwhelms other ling knowledge, impacting LM behavior. We show how to mitigate w/ fine-tuning 🧵



Marten van Schijndel reposted

Curious about how well neural language models learn the same linguistic process in multiple languages? Stop by my poster presentation with @marty_with_an_e at poster session 1I (11am EST) Underline: underline.io/events/167/pos… Paper: aclanthology.org/2021.acl-long.… A very short 🧵 1/4


Marten van Schijndel reposted

Hi all! A 🧵about my #ACL2021NLP paper on discovering complex dynamics in human cognition, which I'll be taking questions about live on Tuesday Aug 3 @ 12:30 EST in ACL session 11B. aclanthology.org/2021.acl-long.… 1/


Loading...

Something went wrong.


Something went wrong.