Gasper Begus

Artificial Humanities: A Fictional Perspective on Language in AI Artificial Humanities: A Fictional Perspective on Language in AI [Beguš, Nina] on Amazon.com. *FREE* shipping on qualifying offers. Artificial Humanities: A Fictional Perspective on Language in AI

Preorder:

www.amazon.com/Artificial-H...

AI is reaching the level of complexity where quantitative evaluations are often not feasible anymore.

The Artificial Humanities book outlines how humanistic methodology can be used to approach and evaluate AI today.

The decade of the humanities is here!

@ninabegus.bsky.social

Check out this amazing plenary talk by Stephan Meylan, a research scientist in our lab:

Modeling Language Acquisition as Speech Development

youtu.be/VP-tUVqmH8U&...

Check out this amazing plenary talk by Stephan Meylan, a research scientist in our lab:

Modeling Language Acquisition as Speech Development

youtu.be/VP-tUVqmH8U&...

Exciting talk in the linguistics department at UC Berkeley tomorrow!
@rtommccoy.bsky.social

Our work on how to model language in humans, animals, and machines with more realistic deep learning approaches in the Communications of ACM today.
cacm.acm.org/news/can-ai-...

Our work on how to model language in humans, animals, and machines with more realistic deep learning approaches in the Communications of ACM today.
cacm.acm.org/news/can-ai-...

#DeepPhonology special session at AMP:

Bruno Ferenc Šegedin

"Evaluating Wasserstein GAN discriminators as models of human well-formedness judgments"

The plenary talk at the hashtag#DeepPhonology Special Session of the AMP 2025 on modeling language with no text from raw data.

"Modeling language acquisition as speech development"

Stephan Meylan (UC Berkeley)

Traditional approaches to language acquisition have conceived of the

The decade of the humanities is coming!

I'll conclude by sketching the potential for rapid progress when we combine rich computational models of individual language processing with linguists' detailed characterizations of language structure.

I'll show how this approach clarifies the communicative burden of adult listeners, the criteria for "successful" language productions, and the utility of videos of speakers' faces for phonological category induction.

and understanding. I'll argue that treating the input and output of the child learner as raw signal streams (rather than symbols) yields many benefits in the treatment of speaker variability, prosody, multi-modal communication, and offers continuity on both ontogenetic and phylogenetic timescales.

In this talk, I will outline an alternative approach (and corroborating deep neural network models) that treat the development of linguistic representations – especially phonological representations — as emergent solutions to the inferential challenges of real-time language production

emergence of speech as a secondary, auxiliary process with respect to the acquisition of an abstract, highly structured set of linguistic rules (e.g., Chomsky's competence-performance distinction; see also de Saussure's langue and parole).

The plenary talk at the hashtag#DeepPhonology Special Session of the AMP 2025 on modeling language with no text from raw data.

"Modeling language acquisition as speech development"

Stephan Meylan (UC Berkeley)

Traditional approaches to language acquisition have conceived of the

How much morphology and phonology does wav2vec 2.0 learns?

"Emergent morpho-phonological patterning in a model of spoken word recognition"
Jon Gauthier (UCSF), Matthew Leonard (UCSF), Canaan Breiss (USC), Edward Chang (UCSF)

Continuing with a talk on phrasal stress and LLMs:
"Can large language models predict English phrasal stress?"

Jinyoung Jo (Stanford), Sean Choi (Santa Clara U), Arto Anttila (Stanford)

My avid reader, at almost 10 years-old: “Big boring book.”

They’ve pretty much grown together!

The decade of the humanities is coming!

Second talk on using deep learning for low resource ASR:

"Comparing Phonological Feature Sets for Low-Resource ASR"

Alessio Tosolini (Yale), Massimo Daul (NYU), Ayla Karakaş (Yale), Claire Bowern (Yale)

First talk on using deep learning for automatic IPA (AutoIPA):

"AI-assisted analysis of phonological variation in English" Virginia Partridge (UMass Amherst), Joe Pater (UMass Amherst), Parth Bhangla (UMass Amherst), Ali Nirheche (UMass Amherst), Brandon Prickett (UMass Amherst)

A special session on Deep Phonology at AMP 2025 is off to a great start.

Our academic predecessors could only dream about modeling phonology from raw data and creating an artificial baby language learner that learns language from sound. Many of these things are now possible.

10 years after the initial idea, Artificial Humanities is here! Thanks so much to all who have preordered it. I hope you enjoy reading it and find this research approach as generative as I do. More to come!

One of the main linguistic conferences on phonology has just started.

It's such a special moment, because who would have thought that a poster on whale phonology would be presented at the Annual Meeting on Phonology.

@projectceti.bsky.social

✨🐳✨👍😊✨

It's great to see how linguistics can help us understand the communication systems of non-human animals.

One of the main linguistic conferences on phonology has just started.

It's such a special moment, because who would have thought that a poster on whale phonology would be presented at the Annual Meeting on Phonology.

@projectceti.bsky.social

How to model learning of Mandarin tones with no supervision, from raw sound, with a realistic model of human language learning?
The model learns the four tones (male only near perfectly) and also replicates stages of tone learning in language acquisition. What is more difficult to learn for children

Deep tonology!