What have neural networks achieved?

[Otto Schnurr-A11505]

Michael Arbib wrote:
> b) What are the "big success stories" (i.e., of the kind the general public
> could understand) for neural networks contributing to the construction of
> "artificial" brains, i.e., successfully fielded applications of NN hardware
> and software that have had a major commercial or other impact?
>
>  *********************************
>  Michael A. Arbib
>  USC Brain Project
>  University of Southern California
>  Los Angeles, CA 90089-2520, USA
>  arbib at pollux.usc.edu
 
 
While our application does not address brain function, it does
represent a successful example of how neural networks are able learn
and synthesize human behavior.
 
Motorola has developed a text-to-speech synthesizer that utilizes
multiple cooperating neural networks, each specializing in a
particular area of human language ability.  This use of neural
networks for both linguistic and acoustic processing produces speech
with exceptional naturalness.  Speech produced by our system has been
found to be more acceptable to listeners than that of other commercial
systems [11].
 
The system excels in learning the specific characteristics of a given
speaker and allows us to develop new dialects and languages rapidly
when compared to other methods.
 
To date, we have developed four voices: two male speakers of American
English, one female speaker of American English and one male speaker
of Mandarin.  Additional linguistic processing has also produced
speech in Spanish, Greek and Turkish with an American accent.
 
 
Regards,
 
Otto Schnurr
 
Speech Processing Research Lab
Chicago Corporate Research Laboratories
Motorola
schnurr at ccrl.mot.com
 
--
 
  [1] Karaali, O., Corrigan, G., Massey, N., Miller, C., Schnurr, O.,
      & Mackie, A. (1998).  A High Quality Text-To-Speech System Composed
      of Multiple Neural Networks.  International Conference on
      Acoustics, Speech and Signal Processing.  Seattle.
 
  [2] Miller, Corey (to appear).  Individuation of Postlexical
      Phonology for Speech Synthesis.  ESCA/COCOSDA Third
      International Workshop on Speech Synthesis, Jenolan Caves
      Australia.
 
  [3] Miller, Corey (1998).  Pronunciation Modeling in Speech Synthesis.
      Doctoral dissertation, University of Pennsylvania.  Philadelphia,
      Pennsylvania.  Published as Technical Report 98-09, Institute for
      Research in Cognitive Science, University of Pennsylvania.
 
  [4] Miller, C., Karaali, O., Massey, N., (1998).  Learning
      Postlexical Variation in an Individual.  Paper presented at the
      Linguistics Society of America Annual Meeting, New York.
 
  [5] Miller, C., Massey, N., Karaali, O. (1998).  Exploring the Nature
      of Postlexical Processes.  Paper presented at the Penn Linguistics
      Colloquium.
 
  [6] Corrigan, G., Massey, N., & Karaali, O. (1997).  Generating Segment
      Durations in a Text-to-Speech System: A Hybrid Rule-Based/Neural
      Network Approach.  In Proceedings of Eurospeech '97.
      pp. 2675-2678.  Rhodes, Greece.
 
  [7] Karaali, O., Corrigan, G., Gerson, I., & Massey, N., (1997).
      Text-to-Speech Conversion with Neural Networks: A Recurrent TDNN
      Approach.  In Proceedings of Eurospeech '97.  pp. 561-564.  Rhodes,
      Greece.
 
  [8] Miller, C., Karaali, O., & Massey, N. (1997).  Variation and
      Synthetic Speech.  Paper presented at NWAVE 26, Quebec, Canada.
 
  [9] Gerson, I., Karaali, O., Corrigan, G., & Massey, N. (1996).  Neural
      Network Speech Synthesis.  Speech Science and Technology (SST-96).
      Australia.
 
 [10] Karaali, O., Corrigan, G., & Gerson, I. (1996).  Speech Synthesis
      with Neural Networks.  Invited paper, World Congress on Neural
      Networks (WCNN-96).  pp. 40-50.  San Diego.
 
 [11] Nusbaum, H., & Luks, T. (1995).  Comparative Evaluation of the
      Quality of Synthetic Speech Produced at Motorola.  Technical Report
      1, University of Chicago.  Chicago, Illinois.