Joseph (Yossi) Keshet

 

Research Assistant Professor

TTI-Chicago

About Me


Since 2009 I have been a Research Assistant Professor at  TTI-Chicago, a philanthropically endowed academic computer science institute located on the University of Chicago campus. Previously I was a postdoctoral research scientist at IDIAP Research Institute, Switzerland, where I worked with Samy Bengio and Hynek Hermansky. I completed my Ph.D. at The School of Computer Science and Engineering at The Hebrew University under the supervision of Yoram Singer in 2007.


My research focuses on developing and analyzing statistical machine learning algorithms for complex and structured tasks. Much of my research is driven by problems in speech and language processing, where typically each problem has a very large number of training examples, is highly structured, and has a unique measure of performance. 


My CV can be downloaded here (PDF).

News

  1. • I am co-organizing the Symposium on Machine Learning in Speech and Language Processing co-located with ICML in Bellevue, Washington on June 27, 2011 with Geoffrey Zweig, Dan Roth and Hal Daume

  2. • I am organizing Illinois Speech Day 2011 at TTI-Chicago on May 15, 2011

  3. • I founded the ISCA special interest group on machine learning for speech and language (SIGML)

  4. • I am in the program committee of 2011 Speech Processing Conference, at Tel-Aviv Academy, Israel

Office

TTI-Chicago, Room 533

6045 S Kenwood Ave,

Chicago, IL 60637

Phone: (773) 834 6850

Fax: (773) 834-9881

Recent Papers


  1. • David McAllester and Joseph Keshet, Generalization Bounds and Consistency for Latent Structural Probit and Ramp Loss, The 25th Annual Conference on Neural Information Processing Systems (NIPS), 2011 (full oral presentation).

  2. • Joseph Keshet, Chih-Chieh Cheng, Mark Stoehr, and David McAllester, Direct Error Rate Minimization of Hidden Markov Models, The 12th Annual Conference of the International Speech Communication Association (Interspeech), Florence, Italy, 2011.

  3. • Andrew Cotter, Nathan Srebro, and Joseph Keshet, A GPU-Tailored Approach for Training Kernelized SVMs, The 17th ACM Conference on Knowledge Discovery and Data Mining (KDD), San Diego, CA, 2011.

  4. • Joseph Keshet, David McAllester, and Tamir Hazan, PAC-Bayesian Approach for Minimization of Phoneme Error Rate, IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Prague, Czech Republic, 2011.

Email: jkeshet@ttic.edu