Artificial Intelligence

1
Artificial Intelligence

• Part III
• By Sherwin Crown
• Class CS4900

2
Microsofts Universal Translator

• Web-based system capable of accurately and
  automatically translating the entire Microsoft
  Product Support Services (PSS) Knowledge Base
  from English into Spanish, providing real-time
  responses to Spanish queries.
• The PSS Knowledge Base is an enormous collection
  of information used to identify and solve
  problems with Microsoft software.

3

• It took a two-decade plus research effort,
  produced an NLP system known internally at
  Microsoft as NLPWin.
• Started out English to Spanish, and after the
  beta testing they are now working on English to
  Japanese, German, French, and Chinese versions.
• These systems will likely save the company
  millions of dollars and that is just the
  beginning.

4
People Involved

• MIT in 1949
• UCLA
• GeorgeTown in 1953
• The Mormon Connection in 1970s
• The IBM Connection in 1983
• Microsoft in 1991

5
WinNLP Framework
6
MIND

• Joseph Pentheroudakis designed the Morphology
  component and the Microsoft Natural Language
  Dictionary (MIND), which was originally built
  using two different machine-readable
  dictionaries, the Longman Dictionary of
  Contemporary English and the American Heritage
  Dictionary.

7
G Component

• George Heidorn spearheaded the development of a
  programming language called G (short for
  Gamma, and also for Grammar or George ).
• G has a lot in common with the AI language LISP,
  except that it includes specialized structures
  for representing the linguistic relationships.

8
MindNet
9
Creation Of MT Using WinNLP

• The creation of the NLPWin Machine Translation
  system takes place in two stages training and
  runtime.
• Training
• The system begins with a pair of equivalent
  sample sentences from a database.
• This is done for 350,000 sentence pairs in
  English and Spanish, applying both heuristics
  (rules) and statistics to find bits of structural
  alignment across the language boundary.

10
Training Diagram
11

• Syntactic paraphrasing may also come into play.
• MindMelding relies on MindNets path-finding and
  lexical similarity routines.
• MindMeld, refers to a highly sophisticated
  process that has consumed the NLP Groups
  research efforts since 1997.

12
Runtime Diagram
13
Where Is It Going?

• Some in the NLP Group envision a time when a
  mega-translator, based on their technology,
  will allow Internet users to converse in
  unrestricted domains instantaneously.
• Unleashing this type of communication power for
  public use could open a entirely new world of
  global interactions.

14

• Any Questions?