Lightnews — Scholar-powered news

Reposted by Grzegorz Chrupała

Marianne de Heer Kloots @mdhk.net · Aug 19

Had such a great time presenting our tutorial on Interpretability Techniques for Speech Models at #Interspeech2025! 🔍

For anyone looking for an introduction to the topic, we've now uploaded all materials to the website: interpretingdl.github.io/speech-inter...

2 14 39

Grzegorz Chrupała @grzegorz.chrupala.me · Aug 18

Roger Moore reminds Interspeech audience that speech is not audible text. Text is a technology.

1 14

Grzegorz Chrupała @grzegorz.chrupala.me · Aug 17

And that's the whole team behind the tutorial on interpretability techniques for speech models at IS2025.

Left to right Charlotte Pouw, Willem Zuidema, Marianne de Heer Kloots, Martijn Bentum, Hosein Mohebbi, Grzegorz Chrupała, Gaofei Shen & Tom Lentz.

1 17

Grzegorz Chrupała @grzegorz.chrupala.me · Aug 17

Wrap-up of the Interspeech tutorial on interpretability for speech models by @mdhk.net

1

Reposted by Grzegorz Chrupała

Gaofei Shen @gaofeishen.com · Jun 13

Come join our tutorial in Rotterdam!

Marianne de Heer Kloots @mdhk.net · Jun 13

The @interspeech.bsky.social early registration deadline is coming up in a few days!

Want to learn how to analyze the inner workings of speech processing models? 🔍 Check out the programme for our tutorial:
interpretingdl.github.io/speech-inter... & sign up through the conference registration form!

Interpretability Techniques for Speech Models — Tutorial @ Interspeech 2025

interpretingdl.github.io

1 7

Grzegorz Chrupała @grzegorz.chrupala.me · May 28

Basque and Catalan don't have an army but they do each have a police force.

2

Grzegorz Chrupała @grzegorz.chrupala.me · May 26

Feature attribution is always tricky. We show just how tricky it is for speech models.

Gaofei Shen @gaofeishen.com · May 26

I am excited to announce that my paper "On the reliability of feature attribution methods for speech classification" has been accepted to #Interspeech2025!
Co-authors: @hmohebbi.bsky.social, Arianna Bisazza, Afra Alishahi, @grzegorz.chrupala.me
Find the preprint here: arxiv.org/abs/2505.16406

On the reliability of feature attribution methods for speech classification

As the capabilities of large-scale pre-trained models evolve, understanding the determinants of their outputs becomes more important. Feature attribution aims to reveal which parts of the input elemen...

arxiv.org

1 4

Reposted by Grzegorz Chrupała

INTERSPEECH 2025 @interspeech.bsky.social · May 20

The notifications are out and we have accepted a record-breaking number of papers! Of the 2332 papers sent for review, 1163 were accepted! Acceptance rate: 49.87%

⚠️ Camera-ready deadline: May 31!
🗓️ Registration opens soon
📨 Subscribe to our newsletter www.interspeech2025.org/newsletter

1 3 9

Grzegorz Chrupała @grzegorz.chrupala.me · May 22

Maybe it found you, as it's been following you all this time...

1 1

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 17

Looking forward!

wimpouw @wimpouw.bsky.social · Apr 17

Happy to share that in a few weeks ill move to Tilburg University as an Associate Professor at the Department of Cognitive Science and Artificial Intelligence
www.tilburguniversity.edu/about/school... Super excited about this! I want to thank Donders/Radboud and many other people for their support!

Department of Cognitive Science and Artificial Intelligence | Tilburg University

Members of the Cognitive Science & Artificial Intelligence (CSAI) department conduct research in the interdisciplinary domain of Artificial Intelligence and Cognitive Science and are involved in teach...

www.tilburguniversity.edu

2

Reposted by Grzegorz Chrupała

Yevgen Matusevych @yevgenm.bsky.social · Apr 17

Last week to apply for our PhD position!

Yevgen Matusevych @yevgenm.bsky.social · Mar 26

Apply for our PhD position in language acquisition / computational linguistics in Groningen until 24 April! Job ad is here:
www.rug.nl/about-ug/wor...

1 2

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 14

Took me a while.

1

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 14

Someone tell them they can use their own domain.

1

Reposted by Grzegorz Chrupała

David Schindler @david-schindler.de · Apr 14

@tilburg-university.bsky.social is on BSky. Let's get them to be more active and actually post something!

1 1 1

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 11

Gobineau gave us the term Aryan, which does indeed have a linguistic connection. But it postdates and is distinct from Caucasian.

1

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 11

The use of Caucasian as a racial category is due to Johann Friedrich Blumenbach and based on skull shape and physical appearance. I wasn't aware he or others made any connections to language families at the time. The whole idea of the Indo-European language family came later I believe.

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 11

We all agree it's bullshit, but I've just never heard the version of the bullshit that claimed Caucasians speak Indo European languages. Even in the 18th century.

1

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 11

Actual Caucasians mostly speak languages unrelated to Indo-European such as Georgian, Chechen, etc.

2 2

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 11

What's the connection to language families?

1

Reposted by Grzegorz Chrupała

Paweł Ausir Dembowski @ausir.bsky.social · Apr 10

I’m again fucking begging anglo media to stop using „Caucasian” to mean „white” ESPECIALLY in any context involving Russia where oppression of actual Caucasians, as in people from ethnic groups inhabiting the Caucasus, is a real ongoing issue.

The Guardian screenshot: „One video, which has had hundreds of thousands of views across different platforms, appears to be a Russian recruitment ad with Chinese subtitles overlaid. It shows Caucasian men leaving their day jobs to fight and asks viewers: "Do you want to show strength here? Is this the path that you long for? You are a tough man, be like them!”

6 51 130

Grzegorz Chrupała @grzegorz.chrupala.me · Apr 10

We listened to speeches and marched through the streets. There were free fries 🍟. Tilburgers were curious and friendly. Protests are not really a thing in the Netherlands 😁

Photo of the crowd demonstrating in Tilburg.

1 5

Reposted by Grzegorz Chrupała

Katya Ivanova 🟥 @katyaoivanova.bsky.social · Apr 10

🟥 Today we #strike! The Dutch govt - w/ support of #CDA & #CU - has chosen to break its promise of structural investments in universities and cut €1B+ from the budget.

Full day of activities: Check out the teach-outs & join the main protest at 12.30, Willemsplein, Tilburg!

@woinactie.bsky.social

11 34

Reposted by Grzegorz Chrupała

Bruno Nicenboim @bruno-nicenboim.fediscience.org.ap.brid.gy · Apr 9

My R package is in RopenSci and CRAN!🚨 🚨 🚨 🚨

*pangoling*: Access to Large Language Model Predictions

This package is mostly for psycholinguists who just want to get the predictability or surprisal values of words without too much fuss (and without using […]

[Original post on fediscience.org]

3 4

Reposted by Grzegorz Chrupała

Arianna Bisazza @arianna-bis.bsky.social · Apr 8

Modern LLMs "speak" hundreds of languages... but do they really?
Multilinguality claims are often based on downstream tasks like QA & MT, while *formal* linguistic competence remains hard to gauge in lots of languages

Meet MultiBLiMP!
(joint work w/ @jumelet.bsky.social & @weissweiler.bsky.social)

Jaap Jumelet @jumelet.bsky.social · Apr 7

✨New paper ✨

Introducing 🌍MultiBLiMP 1.0: A Massively Multilingual Benchmark of Minimal Pairs for Subject-Verb Agreement, covering 101 languages!

We present over 125,000 minimal pairs and evaluate 17 LLMs, finding that support is still lacking for many languages.

🧵⬇️

2 6 21

Reposted by Grzegorz Chrupała

Neil Cohn @neilcohn.bsky.social · Apr 7

I'm now searching for new PhD students who are interested in studying how people's brains understand sequences of images like in comics. So, a mix of comics, cogneuro (EEG), and psycholinguistics (especially syntax/semantics). Please circulate to all who might be interested! tiu.nu/22739

Job opening: Two PhD positions for the neurocognition of the grammar of visual narratives (22739)

tiu.nu

5 57 66