Pinar Donmez, PhD Candidate Language Technologies Institute, School of Computer Science, Carnegie Mellon University 5000 Forbes Ave., Gates Hillman Complex Pittsburgh, PA 15213 USA Office: 6414 Gates Hillman Complex Phone: 412-251-8270 email: pinard AT cs DOT cmu DOT edu
		home | research | publications | projects |teaching

Research

• My research interests are proactive learning, active learning, rank learning, semi-supervised learning, and information retrieval in general.


• I mainly work on proactive and active learning algorithms and their applications in different areas including classification, and rank learning.


• Recently, I have become interested in unsupervised and accurate learning for classification in the absence of gold standard labels.


• In the summer of 2008, I was a research intern at the Text Mining, Search and Navigation group at Microsoft Research. My mentors were Krysta Svore and Chris Burges.

Awards

• I have been a recipient of the Yahoo Key Scientific Challenges Award Program 2009-2010 in the area of Machine Learning.
  Find out more at CMU NewsLetter and Yahoo Honors Future Thought Leaders


• I have been honored with the Best Presentation Award at the Student Research Symposium at LTI 2007

Book Chapter

Donmez, P., Carbonell, G.: From Active to Proactive Learning Methods, to appear as Book Chapter in Recent Advances in Machine Learning
Eds. J. Koronacki, S.T. Wierzchon, Z. Ras and J. Kacprzyk., Springer: Studies in Computational Intelligence, 2009

Publications

• Unsupervised Supervised Learning II: Training Margin Based Classifiers without Labels
  Donmez, P., Balasubramanian, K., Lebanon, G., ArXiv report 1003.0470, 2010
  
• A Probabilistic Framework to Learn from Multiple Annotators with Time-Varying Accuracy
  Donmez, P., Carbonell, J., Schneider, J., to appear in the Proceedings of SIAM Conference on Data Mining (SDM '10), Columbus, 2010
  
• Pairwise Document Classification for Relevance Feedback
  Elsas, J., Donmez, P., Callan, J., Carbonell, J., TREC 2009
  
• Unsupervised Supervised Learning I: Estimating Classification and Regression Errors without Labels
  Donmez, P., Lebanon, G., Balasubramanian, K., to appear in Journal of Machine Learning Research 11, 2010
  
• Unsupervised Estimation of Classification and Regression Error Rates
  Donmez, P., Lebanon, G., Balasubramanian, K., CMU LTI Tech Report CMU-LTI-09-15 (an earlier version of the JMLR paper)
  
• On the Local Optimality of LambdaRank
  Donmez, P., Svore, K., Burges, C.J.C., in SIGIR '09, Boston, USA 2009
  
• Efficiently Learning the Accuracy of Labeling Sources for Selective Sampling
  Donmez, P., Carbonell, J,G., Schneider, J., in KDD '09, Paris, France 2009
  
• Active Sampling for Rank Learning via Optimizing the Area Under the ROC Curve
  Donmez, P., Carbonell, J.G., in ECIR '09, Toulouse, France 2009
  
• Proactive Learning: Cost-Sensitive Active Learning with Multiple Imperfect Oracles
  Donmez, P., Carbonell, J.G., in CIKM '08, Napa Valley, USA 2008
  
• Optimizing Estimated Loss Reduction for Active Sampling in Rank Learning
  Donmez, P., Carbonell, J.G., in ICML '08, Helsinki, Finland 2008
  
• Paired Sampling in Density-Sensitive Active Learning
  Donmez, P., Carbonell, J.G., in ISAIM '08, Florida USA 2008
  
• Dual Strategy Active Learning
  Donmez, P., Carbonell, J. G., Bennett, P.N., in ECML '07, Warsaw Poland 2007
  The official published version of this article is available at Springer-Verlag's website and is © 2007 Springer-Verlag Berlin Heidelberg.
  The detailed empirical results on this paper can be found here
  
• Supporting CSCL with Automatic Corpus Analysis Technology
  Donmez, P., Rose C.P., Stegmann, K., Weinberger, A., Fischer, F., in Proceedings of Computer Supported Collaborative Learning, Taipei, Taiwan 2005,
• Automatic and Semi-Automatic Skill Coding with a View Towards Supporting Online Assessment
  Rose, C., Donmez, P., Gweon, G., Knight, A., Junker, B., Cohen, W., Koedinger, K., Heffernan, N., in Proceedings of AI in Education, Amsterdam 2005
• Oflazer, K., Donmez, P.: SmartReader: An NLP-based Interactive Reading Application for Language Learning, in Lecture Notes in Computer Science, Proceedings of ISCIS'04}, 2004

Projects

• Proactive Learning with Multiple Imperfect Oracles
• Active Sampling in Rank Learning Applications
• Density Sensitive Sampling Methods for Active Learning for Classification
• Ensemble Methods in Active Learning for Classification

Teaching

• Teaching Assistant for MSIT e-Business Course, Spring 2007 and Spring 2008

• Teaching Assistant for Inventing the Future of Services, Fall 2009