Joseph (Yossi) Keshet



Department of Computer Science
Bar Ilan University,
Room 44, Bldg 109
Ramat Gan, 52900, Israel

Email: jkeshet at cs.biu.ac.il

Tel: +972-3-738-4378

I am a professor at the Department of Computer Science at Bar-Ilan University. My research interests concern both machine learning and computational study of human speech and language. In machine learning my research is focused on deep learning and structured prediction, while my research on speech and language is focused on speech processing, speech recognition, acoustic phonetics, and pathological speech.

My technological goal is to improve the state-of-the-art in applications such as automatic speech recognition, speech indexing and retrieval, acoustic scene analysis, and language understanding. My scientific goal is to contribute to research in human speech communication, phonetics, and medical speech pathology using data-driven methods. I believe that exploiting the structure of language and designing theoretically well-founded statistical machine learning algorithms for particular tasks that are able to make use of large datasets, can solve the complex problems involved in speech and language research. To a great extent, my research interests focus on interdisciplinary areas combining the fields of speech science, machine learning, and linguistics. I therefore constantly collaborate with colleagues from those fields.


News



Speech, Language and Deep Learning Lab

The research in the lab is focused on statistical and machine learning techniques applied to the modeling and processing of speech and language. A typical problem in speech and language processing has a very large number of training examples, is sequential, highly structured, and has a unique measure of performance. The lab's goal is to develop rigorous statistical and machine learning algorithms that maximize performance by matching the internal structure of the problem and by optimizing its unique measure of performance.


Felix Kreuk
Ph.D. candidate
New loss functions for training; Adversarial examples; Signal and speech processing

Tzeviya Fuchs
Ph.D. candidate
Spoken term detection; Keyword spotting


Yosi Shrem
Ph.D. candidate
Adversarial training; Foreign-accent conversion

Yael Segal
Ph.D. candidate
Algorithms for prediction and localization of speech objects

Gabi Shalev
Ph.D. candidate
Deep learning and out-of-distribution detection; scaled prediction

Gal Lev
M.Sc. candidate
Scaled predictions using cosine similarity

Bronya Roni Cherniak
M.Sc. candidate
Certified adversarial robustness using perturbations


Talia Ben-Simon
M.Sc. candidate
Trasforming children voices for speech therapy

Tamar Fenster
M.Sc. candidate
Pathological voices through lingustic lens




Alumni


Danny Karmon, M.Sc.
Risk Minimization in Structured Prediction using Orbit Loss

Amir Gottlieb, M.Sc.
Automatic Analysis of Doppler Echocardiography using Structured Prediction

Stav Buchsbaum, M.Sc.
Multiclass Support Vector Machine with a Reject Option

Einat Naaman, M.Sc.
Learning Similarity Functions for Pronunciation Variations


Shua Dissen, M.Sc.
Formant Estimation and Tracking using Deep Networks

Yaniv Sheena, M.Sc.
Speech Segmentation using Deep Structured Models

Yossi Adi, Ph.D.
Deep learning, structured prediction, speech and audio processing


Resources and Code   

The lab is commited to reproducible results. The GitHub repository gives you access to our code, tools and information on how to setup and use. {Deep} Phonetic Tools is a project done in collaboration with Matt Goldrick and Emily Cibelli, where we proposed a set of phonetic tools for measureing VOT, voswel duration, word duration and formants, and are all based on deep learning.



Publications



Copyright disclaimer

This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.