Interaction%20between%20Academia%20and%20Microsoft%20in%20Speech%20and%20Language%20Systems

1
Interaction betweenAcademia and Microsoft in
Speech and Language Systems

• Kentaro Toyama
• Microsoft Research
• Chandar Sundaram, Andy Abbar,
• Alex Acero, Mythreyee Ganapathy, Raveesh Gupta
• SPLASH 2004

2
(No Transcript)
3
Interaction betweenAcademia and Microsoft in
Speech and Language Systems

• Kentaro Toyama
• Microsoft Research
• Chandar Sundaram, Andy Abbar,
• Alex Acero, Mythreyee Ganapathy, Raveesh Gupta
• SPLASH 2004

4
Academia and Microsoft

• Points of interaction for speech language
• Academic Developer Program
• MSDN AA -- Speech SDK
• Future Activities
• Localization
• Local Language Program
• Microsoft Research
• Natural Language Processing
• Speech
• University Relations

5
Academic Developer Program
6
MSDN Academic Alliance

• Subscription to MSDN valid for entire department
• Curriculum Tools
• Over 800 hours of curriculum materials
• Submit curricula youve developed
• Speech SDK bundled with MSDN

7
Future Offerings

• Visual Studio 2005 Express Edition
• The Spoke
• MSDN Academic Sessions
• Imagine Cup 2005
• Project Portal

8
http//msdn.microsoft.com/academic/
9
Localization
10
Localization
Click
Computer

• Challenges
• Time
• 2 languages per year
• Complexity
• People linguists, computer users, developers,
  political scientists, translators, regional
  experts, etc.
• Technology keyboard drivers, character set
  standardization, fonts, currency symbols,
  glossary, translation of help files, etc.
• Cost
• Business case not always present
• Customer involvement
• Microsoft should not unilaterally determine
  computer terminology

11
Local Language Program

• The Local Language Program
• 40 languages per year
• Localizes UI for Windows and Office
• In India
• Hindi is done
• Telegu, Tamil, Kannada, Gujarati on its way
• Resulting glossary, Language Interface Pack
  available to public at no cost
• Involvement by governments and universities
  critical!

12
http//members.microsoft.com/wincg/
13
Microsoft Research
14
Microsoft Research

• Founded in 1991
• Staff of over 650 in over 50 areas
• Internationally recognized research teams
• 5 lab locations around the world
• Research groups in
• Natural Language Processing
• Speech
• Machine Learning
• Search
• Separate University Relations group
• India Mythreyee Ganapathy (mythg_at_microsoft.com)

15
Impact on Product

• Text-to-speech engine (Windows)
• Command and Control (Windows)
• Smart Tags (Office)
• Grammar Checker (Office)
• IntelliShrink text compression (Office)
• Dictation (Office)
• Mandarin Chinese data entry (Office)
• Spam filter (MSN/Exchange)
• Speech API (SAPI)
• Speech Server

16
Research Philosophy

• University organizational model
• Flat structure, critical mass groups
• Open research environment
• Publications strongly encouraged
• Conference attendance high
• Daily lectures by visiting researchers
• Support for university research
• Nearly 15 of basic research budget directly
  invested in universities
• Lab grants, research grants, fellowships, etc.
• Internships for students

17
Text to Speech with Prosody

• MSR Asias Mulan project
• The Speech Group in Microsoft Research Asia is
  conducting research in voice technology, such as
  speech recognition, speech synthesis, and
  speech-enabled information search.

18
Source Separation

• Idea Estimate filters h11n and h12n that
  maximize p(z1n??) where ? is a HMM.
• Approximate HMM by a Gaussian Mixture Model with
  LPC parameters gt EM algorithm with a linear set
  of equations

19
Multimodal Map
20
University Relations

• Liaison to universities
• Emphasis on curriculum and research
• Periodic workshops
• Faculty Summit in Redmond (July)
• India UR manager
• Mythreyee Ganapathy (mythg_at_microsoft.com)

21
http//research.microsoft.com
22
Thank you!kentoy_at_microsoft.com