Joseph Keshet

Joseph ‘Yossi’ Keshet

Speech, Language, and Artificial Intelligence.

I am an Associate Professor at the Viterbi Faculty of Electrical and Computer Engineering at the Technion – Israel Institute of Technology, where I direct the Speech, Language, and AI Lab, affiliated with the Signal and Image Processing Lab (SIPL). I am also a Courtesy Faculty Member with the Faculty of Data and Decision Sciences.

I am a Deputy Editor of IEEE Transactions on Audio, Speech, and Language Processing and was selected as a 2026–2027 ISCA Distinguished Lecturer.

Full CV

My Research

I am passionate about human speech—one of the most fundamental yet intricately complex signals we know. Speech carries not only the information a speaker intends to convey but also reveals their identity, emotional state, and even medical condition. My drive to understand and quantify speech shapes my research interests, which span both machine learning and the computational study of human speech and language.

Blog

News

Deputy Editor for IEEE Trans. on Audio, Speech, and Language Processing
Privileged to have been appointed Deputy Editor for IEEE Transactions on Audio, Speech, and Language Processing.
ISCA Distinguished Lecturer
Appointed 2026–2027 ISCA Distinguished Lecturer. Grateful for the opportunity to advance knowledge and support global speech and language research communities.
Workshop
iSpeech 2025
On Sept 17, 2025, we will hold the second Israeli conference on Speech & AI at the Technion. The program and registration can be found here.  
Committee
IEEE Speech and Language Processing Technical Committee
Elected to the IEEE Speech and Language Processing Technical Committee (SLTC).
Interview
Why Amazon Scholar Yossi Keshet remains ‘excited about speech’
A blog post at Amazon Science by Larry Hardesty (click to read).

Publications

Transformer architectures, similar to those behind ChatGPT, are the foundation of today’s most successful speech recognition systems. Likewise, diffusion-based models have become key to state-of-the-art speech synthesis and generation. My recent publications focus on improving the performance and conceptual understanding of these architectures within the speech domain.
Sound

All publications

Open-vocabulary keyword spotting with hyper-matched filters for small footprint devices

Yael Segal-Feldman, Ann R. Bradlow, Matthew Goldrick, Joseph Keshet

Computer Speech & Language, Volume 100, October 2026, 101986.
Joint Enhancement and Classification using Coupled Diffusion Models of Signals and Logits

Gilad Nurko, Roi Benita, Yehoshua Dissen, Tomohiro Nakatani, Marc Delcroix, Shoko Araki, Joseph Keshet

Preprint, 2026.
Analyzing and Guiding Zero-Shot Posterior Sampling in Diffusion Models

Roi Benita, Michael Elad, Joseph Keshet

Preprint, 2026.
How Does a Deep Neural Network Look at Lexical Stress?

Itai Allouche, Itay Asael, Rotem Rousso, Vered Dassa, Ann Bradlow, Seung-Eun Kim, Matthew Goldrick, Joseph Keshet

The Journal of the Acoustical Society of America,  Vol. 159, Issue 2, pp. 1348-1358, 2026.
Large Language Model Guided Decoding for Self-Supervised Speech Recognition

Eyal Cohen, Bhiksha Raj, Joseph Keshet

Preprint, 2026
Beyond Transcription: Mechanistic Interpretability in ASR

Neta Glazer, Yael Segal-Feldman, Hilit Segev, Aviv Shamsian, Asaf Buchnick, Gill Hetz, Ethan Fetaya, Joseph Keshet, Aviv Navon

The 40th Annual AAAI Conference on Artificial Intelligence (AAAI-26), 2026

Courses

Speech Processing with Deep Learning

Speech recognition and synthesis have become foundational components in the development of modern artificial intelligence, bridging the gap between human communication and machine interaction. This course aims to provide a comprehensive introduction to the core principles of speech signal processing and modeling, essential for understanding how machines perceive and generate human language.

Read More

Deep Learning

This course covers key theoretical and practical tools for deep learning, with an emphasis on supervised learning. Topics include gradient descent, fully-connected and convolutional networks, training methods, and architectures for sequential data like transformers. We also explore techniques to improve data and resource efficiency, such as pre-training, self-supervised learning, quantization, and pruning.

Read More

Transformers and Large Language Models

Transformers are powerful deep learning architectures that have revolutionized how machines understand and generate sequential data. From enabling breakthroughs in language models like ChatGPT to powering cutting-edge systems in speech recognition and computer vision, Transformers have become a cornerstone of modern AI. This course explores their design, capabilities, and impact across a range of domains.

Read More
|ECE 460747 / DDS 970201|ECE 460217|ECE 0480011