Computational and Experimental Methods

Beguš speaks at Harvard

September 18, 2024

Gašper Beguš is giving a talk at the linguistics department at Harvard University on Friday, September 20 (https://linguistics.fas.harvard.edu/blog/week-sept-16). The title of the talk is "Building realistic models of language with deep learning". Congrats!

Gašper Beguš

Associate Professor of Linguistics

PhD, Harvard

Phonology, phonetics, computational linguistics, historical linguistics, Indo-European

Galvano and D-Lab at OpenAI Forum's "AI Ethics in Action"

August 27, 2024

This past Wednesday, Amber Galvano and colleagues represented UC Berkeley's D-Lab at the OpenAI Forum's "AI Ethics in Action" virtual event, where they discussed their personal trajectory and experience participating in the Data Science for Social Justice workshop.

Beguš presents two talks

October 26, 2023

Gašper Beguš gave two invited talks this week, one at Stanford NLP Seminar (https://nlp.stanford.edu/seminar/details/gasperbegus_2023.shtml) and the other at the Linguistics Colloquium at UC Davis.

Beguš speaks at U of Arizona

September 26, 2023

Gašper Beguš gave an invited colloquium talk on "Modeling Language as a dependency between the latent space and data" at the University of Arizona's Department of Linguistics on September 22, 2023. The abstract can be found here: https://linguistics.arizona.edu/events/modeling-language-dependency-between-latent-space-and-data

Beguš presents at IIT Guwahati

September 21, 2023

Gašper Beguš gave a virtual invited hall titled Modeling language from raw speech with GANs” at the CHAI: Chat about AI colloquium at the School of Data Science and AI, Indian Institute of Technology (IIT Guwahati) on September 13, 2023.

Beguš on podcasts

September 14, 2023
Gašper Beguš recently appeared on the Stack Overflow Podcast and the STARTS podcast talking about computers and language. You can listen to the podcasts at the following links: https://stackoverflow.blog/2023/09/07/ai-brain-computer-interface-deep-implant-speech/, https://open.spotify.com/episode/2LQEGSrtoXRBUZVAhINdEI

Beguš in ICASSP 2023

May 8, 2023

Gašper Beguš published a paper titled "Articulation GAN: Unsupervised modeling of articulatory learning" in proceedings of ICASSP 2023 (IEEE International Conference on Acoustics, Speech and Signal Processing) with Alan Zhou, Peter Wu, and Gopala K. Anumanchipalli. The paper is available here.

Video of the presentation scheduled to be given at the conference in Rhodes, Greece on June 9 is available here.

Interpreting Intermediate Convolutional Layers of Generative CNNs Trained on Waveforms

Gašper Beguš
Alan Zhou
2022

This paper presents a technique to interpret and visualize intermediate layers in generative CNNs trained on raw speech data in an unsupervised manner. We argue that averaging over feature maps after ReLU activation in each transpose convolutional layer yields interpretable time-series data. This technique allows for acoustic analysis of intermediate layers that parallels the acoustic analysis of human speech data: we can extract F0, intensity, duration, formants, and other acoustic properties from intermediate layers in order to test where and how CNNs encode various types of...