Portrait of Jason Williams

Jason Williams

Principal Researcher

About

Research interests

Dialog management: Applications of machine learning; integration of expert knowledge; multi-modal dialog management; planning techniques; on-line improvement.

Dialog and user modeling: Tracking and quantifying uncertainty in dialog state for human/computer dialog; representational structures for dialog state; ontology integration; simulation.

Turn-taking: Use of continuous/incremental speech recognition for turn-taking; integration of conversation history and user model for turn-taking.

Confidence scoring: Machine-learning-based approaches to confidence scoring; calibration in confidence scoring; features for confidence scoring.

Planning under uncertainty: Markov decision processes (MDPs); partially observable Markov decision processes (POMDPs); reinforcement learning.

Experience

  • Microsoft Research, Senior Researcher, 2012-Present
  • AT&T Labs Research, Principal Member of Technical Staff, 2006-2012
  • Cambridge University, Ph D, Engineering Dept, 2002-2006
  • Edify Corp, Senior consultant – Usability and Speech Technology, 2002-2005
  • Tellme Networks, Voice Application Development Manager, 2000-2001
  • McKinsey & Company, Associate, 1999-2000
  • Cambridge University, Masters, Speech/Language Processing, 1998-1999
  • Princeton University, BSE, Electrical Engineering, 1994-1998

Research awards

  • SigDial best paper award: 2014
  • IEEE SLT Best Poster Presentation in session: 2014
  • SigDial Best student paper award: 2011
  • ISCA award for best paper in Computer Speech and Language, for the period 2005-2009: 2010
  • AT&T Key Contributor award: 2010, 2009
  • AT&T Intellectual Property achievement award: 2010
  • AT&T Labs President Excellence Award: 2010
  • AT&T Labs Research excellence award: 2007
  • AVIOS Best paper award: 2003

Educational materials

  • Jason D. Williams, Spoken dialog systems as an application of planning under uncertainty, in Invited talk at International Conference on Automated Planning and Scheduling (ICAPS), Freiburg, Germany, 2011.
  • Jason D. Williams, Speech technologies for interactive mobile applications – a primer, in Invited talk at Joint workshop of the Association for Voice Interaction Design (AVIxD) and the Interaction Design Association (IxDA), New York City, USA, 2011.
  • Jason D. Williams, Steve Young, and Blaise Thomson, Statistical approaches to dialogue systems (tutorial, including videos), 2009.
  • Jason D. Williams, Blaise Thomson, and Steve Young, Bibliography for Statistical Approaches to Spoken Dialogue Systems , August 2009.
  • Jason D. Williams, Spoken dialogue systems: challenges, and opportunities for research (invited talk), in Proc IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU), Merano, Italy, 2009.

Thesis

  • Jason D. Williams, Partially Observable Markov Decision Processes for Spoken Dialogue Management, Cambridge University Engineering Department, 2006.

Thesis committees

Interns

  • Nobal Niraula, Summer 2014. Natural language understanding. Currently a PhD student with Vasile Rus, University of Memphis.
  • Pradeep Dasigi, Summer 2014. Conversational systems. Currently a PhD student with Ed Hovy, CMU.
  • Ke Zhai, Summer 2013. Dialog modeling. Currently a Research Scientist with Yahoo.
  • He He, Summer 2013. Action selection in dialog systems (co-mentored with Lihong Li as lead mentor). Currently a PhD student with Hal Daumé III at UMD.
  • Angeliki Metallinou, Summer 2012. Belief tracking in dialog systems (co-mentored with Dan Bohus). Currently a Speech Scientist at Amazon Lab126.
  • Ethan Selfridge, Summer 2010 and 2011 at AT&T. Turn-taking in dialog systems. Currently at Interactions Corp.
  • Hamid Chinaei, Summer 2010 at AT&T. Reinforcement learning for dialog systems. Currently at IBM and Ryerson University.
  • John Asmuth, Summer 2009 at AT&T. Bayesian approaches to reinforcement learning. Currently at Google.
  • Lihong Li, Summer 2008 at AT&T. Feature selection in reinforcement learning. Currently a Researcher with Microsoft Research.
  • Umar Syed, Summer 2007 at AT&T. Learning user models from unlabeled data. Currently a research scientist at Google.

Patents

  • 8010364: System and method for applying probability distribution models to dialog systems in the troubleshooting domain
  • 8140328: User intention based on N-best list of recognition hypotheses for utterances in a dialog
  • 8433578: System and method for automatically generating a dialog manager
  • 8457968: System and method for efficient tracking of multiple dialog states with incremental recombination
  • 8473292: System and method for generating user models from transcribed dialogs
  • Plus seven more patent applications in progress with the US Patent Office

Appointments, editor, and organizer roles

Journal reviewing and grant panelist

Conference program committees and reviewing

  • IUI reviewer: 2015, 2014, 2013
  • ICMI reviewer: 2015, 2014
  • CHI reviewer: 2013
  • INTERSPEECH Scientific Review Committee: 2015, 2014, 2013, 2012, 2011
  • IWSDS program committee: 2015, 2014
  • SIGDIAL program committee: 2013, 2012, 2011, 2010, 2009, 2008, 2007
  • ASRU Workshop program committee: 2013, 2011, 2009, 2007
  • EACL scientific review committee: 2014, 2012, 2009
  • ACL Student Research workshop program committee: 2014, 2015
  • NAACL Student Research workshop program committee: 2015
  • ACL program committee: 2015, 2014, 2013, 2012, 2011
  • ACL demo committee: 2011
  • NAACL reviewer: 2012, 2009, 2007
  • IJCAI senior program committee: 2015
  • IJCAI program committee: 2011, 2009, 2007
  • EMNLP program committee: 2015, 2014, 2013, 2009, 2008, 2006
  • NIPS reviewer: 2015, 2010
  • ICASSP reviewer: 2014, 2013, 2012, 2011, 2010, 2009, 2008, 2007
  • EACL 2014 Workshop on Dialog in Motion: 2014
  • SLAM: Joint ISCA/IEEE International Workshop on Speech, Language and Audio in Multimedia: 2013
  • ECAI/IJCAI/AAAI Workshop on Machine Learning for Interactive Systems (MLIS): 2015, 2014, 2013, 2012
  • IJCAI Workshop on Knowledge and Reasoning in Practical Dialogue Systems program committee: 2011, 2009
  • IEEE Workshop on Spoken Language Technology (SLTC) technical program committee: 2014, 2012, 2010, 2008
  • COLING Program committee: 2014
  • COLING Demonstration session: 2014
  • COLING Workshop: Spoken language technologies for pervasive speech-based and multimodal applications program committee: 2014, 2008
  • Young researchers’ roundtable on spoken dialogue systems advisory board: 2015, 2014, 2013, 2012, 2011, 2010, 2009, 2006
  • Workshop on Future directions and needs in the Spoken Dialog Community: Tools and Data scientific committee: 2012
  • EMNLP Workshop on Modeling Large Scale Social Interaction in Massively Open Online Courses: 2014
  • NIPS Workshop on Learning Semantics: 2014
  • European Signal Processing Conference (EUSIPCO): 2015
  • Workshop on Future and Emerging Trends in Language Technology: 2015

Projects

Dialog and Conversational Systems Research

Established: March 14, 2014

Conversational systems interact with people through language to assist, enable, or entertain. Research at Microsoft spans dialogs that use language exclusively, or in conjunctions with additional modalities like gesture; where language is spoken or in text; and in a variety…

Publications

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

Projects

Other

Language Understanding Intelligent Service (LUIS)

LUIS is a new, fast way of building language understanding models. For more information, see projectoxford.ai/luis. LUIS is currently in private beta. To request an invitation, visit luis.ai.

Dialog State Tracking Challenge

The Dialog State Tracking Challenge (DSTC) is a research community challenge task for accurately estimating a user’s goal in a spoken dialog system. DSTC homepage.

Software

AT&T Statistical Dialog Toolkit (ASDT): Software for tracking a distribution over a large number of hidden states in a spoken dialog systems. Available under license agreement from AT&T Labs Research.