Umut Ozertem

Umut Ozertem
SENIOR SCIENCE LEAD
.

Umut is a Senior Science Lead at Microsoft, and he leads a team that works on machine learning and data mining techniques and applies them to Automatic Speech Recognition and Spoken Language Understanding. Massively parallel data processing, large scale text (or query log) mining for statistical semantics, personalization/contextualization of language models, and things like that…

Prior to joining the Speech team at Microsoft, Umut was a member of the Bing whole page relevance team, a Scientist at Yahoo! Labs (2008–2011), a graduate intern at Intel (January–July 2007), and a graduate research assistant at the Oregon Graduate Institute, where he received his M.S. and Ph.D. in EE, in 2006 and 2008.

Umut has been a program committee member or session chair for many conferences (WWW, NIPS, ICASSP, MSLP, ICANN, ICA, EMBC,…) and has been serving as a reviewer to the following journals (and probably some more): IET Signal Processing, SPIE Journal of Electronic Imaging, Journal of VLSI Signal Processing, Signal Processing, Neurocomputing, Pattern Recognition Letters, IEEE’s Trans. on Image Processing, Trans. on Signal Processing, Trans. on Neural Networks, Signal Processing Letters, Trans. on Biomedical Engineering.

Umut is a long time Göztepe fan. He plays guitar, snowboards –and sometimes breaks bones. He is in love with Çağrı.

Patents and Publications

2013

  • Umut Ozertem, Debora Donato, Luca Aiello, “Method and System for Categorizing Web-Search Queries in Semantically Coherent Topics,” US Patent, May 2013

2012

  • Omer Emre Velipasaoglu, Umut Ozertem, “Method and System for Personalized Search Suggestions,” US Patent, November 2012
  • Omer Emre Velipasaoglu, Umut Ozertem, Alpa Jain, “System and Method for Contextualizing Query Instructions Using User's Recent Search History,” US Patent, November 2012
  • Umut Ozertem, Olivier Chapelle, Pinar Donmez, Emre Velipasaoglu, “Learning to Suggest: A Machine Learning Framework for Ranking Query Suggestions,” Proceedings of the 35th ACM SIGIR, pp. 25–34, 2012

2011

  • Omer Emre Velipasaoglu, Alpa Jain, Umut Ozertem, “Synthesized Suggestions for Web-Search Queries,” US Patent, Jan 2011
  • Sihem Amer-Yahia (editor) – with 16 co-authors from Yahoo!, “Recommendation Projects at Yahoo!,” IEEE Data Engineering Bulletin, vol. 34, no. 2, pp. 69–77, 2011
  • Umut Ozertem, Deniz Erdogmus, “Principal Curves and Surfaces Emerge from Local Density Structure,” Journal of Machine Learning Research, vol. 12, pp. 1249–1286, 2011
  • Luca Maria Aiello, Umut Ozertem, Debora Donato, Filippo Menczer, “Behavior-driven Clustering of Queries into Topics“, Proceedings of the 20th ACM CIKM, pp. 1373–1382, 2011
  • Umut Ozertem, Emre Velipasaoglu, Larry Lai “Suggestion Set Utility Maximization Using Session Logs”, Proceedings of the 20th ACM CIKM, pp. 105–114, 2011
  • Alpa Jain, Umut Ozertem, Emre Velipasaoglu, “Synthesizing High Utility Suggestions for Rare Web Search Queries”, Proceedings of the 34th international ACM SIGIR, pp. 805–814, 2011
  • Umut Ozertem, Rosie Jones and Benoit Dumoulin, “Evaluating New Search Engine Congurations with Pre-existing Judgments and Clicks”, Proceedings of the 20th international conference on World Wide Web, pp. 397–406, 2011

2010

  • Gilad Avraham Mishne, Umut Ozertem, Network-Resource-Specific Search Assistance, US Patent, October 2010
  • Hakan Ceylan, Rada Mihalcea, Umut Ozertem, Elena Lloret, Manuel Palomar, “Quantifying the Limits and Success of Extractive Summarization Systems Across Domains”, Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pp. 903–911, 2010

2009

  • Steven V. Vaerenbergh, Ignacio Santamaria, Paolo E. Barbano, Umut Ozertem, Deniz Erdogmus, “Path-based Spectral Clustering for Decoding Fast Time-varying MIMO Channels,” IEEE Workshop on Machine Learning for Signal Processing, pp. 1–6, 2009
  • Jayashree Kalpathy-Cramer, Umut Ozertem, M. Fuss, Deniz Erdogmus, “Robust segmentation using nonparametric snakes with multiple cues for applications in radiation oncology,” SPIE Medical Imaging, vol. 7259, pp. 72594S–72594S-9, 2009
  • Ervin Sejdic, Umut Ozertem, Igor Djurovic, Deniz Erdogmus, “A New Approach for the Reassignment of Time-Frequency Representations,” IEEE Int. Conf. on Acoustics Speech and Signal Processing, pp. 2997–3000, 2009
  • Tamara Hayes, Misha Pavel, Jay Lundell, Farzin Guilak, Sengul Vurgun, Umut Ozertem, Kevin Rhodes, Kofi Cobbinah, “A Study of Medication-Taking Adherence and Unobtrusive, Intelligent Reminding”, Telemedicine and E-health, vol. 15, no. 8, pp.770–776, 2009
  • Umut Ozertem, Deniz Erdogmus, “RKHS Bayes Discriminant: A Subspace Constrained Detection Filter,” IEEE Trans. on Neural Networks, vol 20, no. 1, pp 1195–1203, 2009
  • Umut Ozertem, Deniz Erdogmus, “Principal Curve Time Warping,” IEEE Trans. on Signal Processing, vol. 57, no. 6, pp.2041–2049, 2009

2008

  • Umut Ozertem, Ismail Uysal, Deniz Erdogmus, “Continuously Differentiable Sample Spacing Entropy Estimation,” IEEE Trans. on Neural Networks, vol. 19, no. 11, pp. 1978–1984, 2008
  • Umut Ozertem, Deniz Erdogmus, “Second Order Volterra System Identification with Noisy Input-Output Measurements,” IEEE Signal Processing Letters, vol. 16, no. 1, pp. 18–21, 2008
  • Deniz Erdogmus, Umut Ozertem, Tian Lan, “Information Theoretic Feature Selection and Projection,” Speech, Audio, Image and Biomedical Signal Processing using Neural Networks, Springer-Verlag, 2008
  • Puskal P. Pokharel, Umut Ozertem, Deniz Erdogmus, Jose C. Principe, “Recursive Complex BSS via Generalized Eigendecomposition and Application in Image Rejection for BPSK,” Signal Processing, vol. 88, no. 6, pp. 1368–1381, 2008
  • Umut Ozertem, Deniz Erdogmus, Robert Jenssen, “Mean Shift Spectral Clustering,” Pattern Recognition, vol. 41, no. 6, pp. 1924–1938, 2008
  • Jayashree Kalphathy-Cramer, Umut Ozertem, William Hersh, Martin Fuss, Deniz Erdogmus, “Semi-supervised Segmentation using Nonparametric Snakes for 3D-CT Applications in Radiation Oncology,” IEEE Int. Conf. on Machine Learning for Signal Processing, pp. 109–114, 2008
  • Umut Ozertem, Deniz Erdogmus “Principal Graphs and Piecewise Linear Subspace Constrained Mean Shift,” IEEE Int. Conf. on Machine Learning for Signal Processing, pp. 438–443, 2008
  • Umut Ozertem, Deniz Erdogmus, Orhan Arikan, “Piecewise Smooth Signal Denoising via Principal Curve Projections,” IEEE Int. Conf on Machine Learning for Signal Processing, pp. 426–431, 2008
  • Erhan Bas, Deniz Erdogmus, Umut Ozertem, Misha Pavel, “Towards sh-eye camera based in-home activity assessment,” IEEE Int. Conf. of Engineering in Medicine and Biology Society, pp. 2558–2561, 2008
  • Umut Ozertem, Deniz Erdogmus, “Signal Denoising Using Principal Curves with an Application to Time Warping,” IEEE Int. Conf. on Acoustics Speech and Signal Processing, pp. 3709–3712, 2008
  • Umut Ozertem, Deniz Erdogmus, Miguel A. Carreira-Perpinan, “Density Geodesics for Similarity Clustering,” IEEE Int. Conf. on Acoustics Speech and Signal Processing, pp. 1977–1980, 2008
  • Umut Ozertem, Deniz Erdogmus, “Local Conditions for Critical and Principal Manifolds,” IEEE Int. Conf. on Acoustics Speech and Signal Processing, pp. 1893–1896, 2008

2007

  • Umut Ozertem, Deniz Erdogmus “Nonparametric Snakes,” IEEE Trans. on Image Processing, vol. 16, no. 9, pp. 2361–2368, 2007.
  • Umut Ozertem, Deniz Erdogmus “Information Regularized Sensor Fusion: Application to Localization with Distributed Motion Sensors,” JVLSI Signal Processing, vol. 49, no. 2, pp.291–299, 2007
  • Deniz Erdogmus, Umut Ozertem, “Nonlinear Coordinate Unfolding via Principal Curve Projections with Application to Nonlinear BSS,” Int. Conf. on Neural Information Processing, pp. 488–497, 2007
  • Jay Lundell, Tamara Hayes, Sengul Vurgun, Umut Ozertem, Janna Kimel, Jeffery A. Kaye, Farzin Guilak, Misha Pavel “Continuous Activity Monitoring and Intelligent Contextual Prompting to Improve Medication Adherence,” IEEE Int. Conf. of Engineering in Medicine and Biology Society, pp. 6286–6289, 2007
  • Umut Ozertem, Deniz Erdogmus, “A Nonparametric Approach for Active Contours,” Proc. of Int. Joint. Conf. on Neural Networks, pp. 1407–1410, 2007
  • Umut Ozertem, Andras Gruber, Deniz Erdogmus, ” Automatic Brain Image Segmentation for Evaluation of Experimental Ishemic Stroke Using Gradient Vector Flow and Kernel Annealing,” Proc. of Int. Joint. Conf. on Neural Networks, pp. 1397–1400, 2007
  • Deniz Erdogmus, Umut Ozertem, “Self-Consistent Locally De ned Principal Surfaces,” Proc. of Int. Conf. on Accoustics, Speech, and Signal Processing, vol. 2, pp. II-549–II-552, 2007
  • Umut Ozertem, Deniz Erdogmus, “Information Regularized Maximum Likelihood for Binary Motion Sensors,” Int. Conf. on Accoustics, Speech, and Signal Processing, vol. 2, pp. II-1021–II-1024, 2007
  • Puskal P. Pokharel, Umut Ozertem, Deniz Erdogmus, Jose C. Principe, “Recursive Complex Blind Source Separation via Eigendecomposition of Cumulant Matrices,” ICASSP, vol. 2, pp. II-645–II-648, 2007

2006

  • Anant Hegde, Deniz Erdogmus, Yadunandana N. Rao, Hemanth Peddaneni, Umut Ozertem, Jose C.Principe, “Perturbation-based Eigenvector Updates for On-line Principal Components Analysis and Canonical Correlation Analysis,” Journal of VLSI Signal Processing Systems, vol. 45, no. 1–2, pp. 85–95, 2006
  • Umut Ozertem, Deniz Erdogmus, Robert Jenssen, “Spectral Feature Projections That Maximize Shannon Mutual Information with Class Labels,” Pattern Recognition, vol. 39, no. 7, pp. 1241–1252, 2006
  • Umut Ozertem, Deniz Erdogmus, “Maximum Entropy Approximation for Kernel Machines,” Proc. of Machine Learning for Signal Processing, pp. 349–352, 2006
  • Deniz Erdogmus, Umut Ozertem, “Clustering with Normalized Information Potential Constrained Maximum Entropy Boltzmann Distribution,” Proc. of Int. Joint Conf. on Neural Networks, pp. 4892–4897, 2006
  • Tian Lan, Deniz Erdogmus, Umut Ozertem, Y. Huang, “Estimating Mutual Information Using Gaussian Mixture Model for Ranking and Selection,” Int. Joint Conf. on Neural Networks, pp. 5034–5039, 2006
  • Deniz Erdogmus, Miguel A. Carreira-Perpinan Umut Ozertem, “Kernel Density Estimation, Affnity-Based Clustering, and Typical Cuts,” Proc. of Int. Conf. on Acoustics, Speech and Signal Processing, vol. 5, pp. V-569–V-572, 2006
  • Umut Ozertem, Deniz Erdogmus, Tian Lan, “Mean Shift Spectral Clustering for Perceptual Image Segmentation,” Int. Conf. on Acoustics, Speech and Signal Processing, vol. 2, pp. II–117–II-120, 2006
  • Umut Ozertem, Deniz Erdogmus, Tian Lan, “Recursive Generalized Eigendecomposition for Independent Components Analysis,” Lecture Notes in Computer Science, vol. 3889, pp. 198–205, 2006

2005

  • Umut Ozertem, Deniz Erdogmus, “Supervised Neural Network Training Using Minimum Error Entropy with Variable-size and Finite-Support Kernel Estimates,” Proc. of Machine Learning for Signal Processing, pp. 67–72, Sep 2005
  • Umut Ozertem, Deniz Erdogmus, “Spectral Clustering with Mean Shift Preprocessing,” Proc. of Machine Learning for Signal Processing, pp. 73–78, Sep 2005
  • Umut Ozertem, Deniz Erdogmus, “Maximally Discriminative Spectral Feature Projections Using Mutual Information,” Proc. of Int. Joint Conf. on Neural Networks, vol. 1, pp. 208–213, Aug. 2005
  • Umut Ozertem, Deniz Erdogmus, Ignacio Santamaria, “Detection of Nonlinearly Distorted Signals Using Mutual Information,” European Signal Processing Conference, 2005

Contact Information

Microsoft Silicon Valley, Sunnyvale CA 94089

umut.ozertem@microsoft.com