[Elsnet-list] announcement elsnet

Louis ten Bosch tenbosch at LandS.let.ru.nl
Wed Dec 21 14:56:48 CET 2005

CALL FOR PAPERS for special issue of Speech Communication

Special Issue on Bridging the Gap Between Human and Automatic Speech

In the past decade the performance of speech processing systems has
improved dramatically, resulting in the widespread use of speech
technology in many real-world applications. State-of-the-art large
vocabulary speech recognition systems are capable of achieving word error
rates of 10-15% or lower on data of the most difficult type (natural,
conversational speech). To a large extent, these achievements are due to
the use of unprecedented amounts of training data. This has led many
speech researchers to believe that future improvements can
be obtained by ever-increasing amounts of training data, as well as
efficient techniques for utilizing these resources.

On the other hand, speech processing performance still falls short of
human performance in many ways, particularly with respect to robustness to
mismatches in training and test conditions. An entirely alternative view
of the current state of speech recognition research is that the remaining
limitations of speech processing can only be solved by radical
modifications to the standard paradigm and, in particular, by integrating
more knowledge about crucial characteristics of human speech perception.

This alternative is particularly interesting since the field of human
speech perception research is increasingly utilizing computational and
statistical techniques, some of which are inspired by the statistical
matching techniques use in the ASR paradigm.

This special issue of Speech Communication is entirely devoted to studies
that seek to bridge the gap between human and automatic speech
recognition. It follows the special session at INTERSPEECH 2005 on the
same topic.

Papers are invited that cover one or several of the following issues:

- quantitative comparisons of human and automatic speech processing
  capabilities, especially under varying environmental conditions

- computational approaches to modelling human speech perception

- use of automatic speech processing as an experimental tool in
  human speech perception research

- speech perception/production-inspired modelling approaches for speech
  recognition, speaker/language recognition, speaker tracking,
  sound source separation

- use of perceptually motivated models for providing rich
  transcriptions of speech signals (i.e. annotations going beyond
  the word, such as emotion, attitude, speaker characteristics, etc.)

- fine phonetic details: how should we envisage the design and evaluation
  of computational models of the relation between fine phonetic details in the
  signal on the one hand, and key effects in (human) speech processing on the
  other hand.

- how can advanced detectors for articulatory-phonetic features be integrated
  in the computational models for human speech processing

- the influence of speaker recognition on speech processing

Guest Editors:
Katrin Kirchhoff                         Louis ten Bosch
Department of Electrical Engineering     Dept. of Language and Speech
University of Washington                 Radboud University Nijmegen
Box 352500                               Post Box 9103
Seattle, WA, 98195                       6500 HD Nijmegen
(206) 616-5494                           +31 24 3616069
katrin at ee.washington.edu                 l.tenbosch at let.ru.nl

Important Dates:
Submission Deadline: April 30, 2006
Notification of Acceptance: August 30, 2006
Final Manuscript Due: September 30, 2006
Tentative Publication Date: December 2006

Submission Procedure:
Prospective authors should follow the regular guidelines of the Speech
Communication Journal for electronic submission
(http://ees.elsevier.com/specom). Authors must select the Section:
Special Issue Paper: Bridging the Gap... , and not Regular Paper
and request Professor Julia Hirschberg as Managing Editor for the paper.

More information about the Elsnet-list mailing list