EUROSPEECH 97
COCOSDA 97 Satellite Event
26 September 1997

EAGLES Spoken Language Working Group
Results and Current Work

Dafydd Gibbon, U Bielefeld
Technical Editor

• EAGLES Spoken Language Group:
  Project page
  
• Publication information:
  `Handbook of Standards and Resources for Spoken Language Systems'

EAGLES I: project structure

EAGLES I: SLWG

• Coordination:
  • Chairman: Roger Moore
  • Host: Richard Winski
  • Editor: Dafydd Gibbon
  • Core group
  • Area subgroups
• SLWG Goal: Production of `Handbook of Standards and Resources for Spoken Language Systems'
• SLWG Policy and Authors: Generic, multilingual methodologies; consensus negotation cycles; experts from 10 European laboratories
• SLWG Coverage:
  System design, SL Corpora (design, recording, processing), SL Lexica, Assessmenti methodology (introductory chapter; speech input, speaker verification, and speech output devices), Appendices on current standards in specific areas.

EAGLES I: SLWG subgroups

EAGLES I: SLWG Handbook Topics

User's guide

Part I: SL System and Corpus Design

• System design
• SL corpus design
• SL corpus collection
• SL corpus representation

Part II: SL characterisation

• Spoken language lexica
• Language models
• Physical characterisation and description

Part III: SL system assessment

• Assessment methodologies and
  experimental design
• Assessment of recognition systems
• Assessment of speaker verification systems
• Assessment of synthesis systems
• Assessment of interactive systems

EAGLES I: Handbook reference materials

• Character codes and computer readable alphabets
• SAMPA computer readable phonetic alphabet
• SAM file formats
• SAM recording protocols
• SAM software tools
• EUROPEC recording tool
• Digital storage media
• Database Management Systems (DBMSs)
• Speech standards
• EUROM-1 database overview
• Polyphone project overview
• European speech resources
• Transcription and documentation conventions for Speechdat
• The Bavarian Archive for Speech Signals

EAGLES I: SLWG main technical authors

User's guide Roger Moore Part I:

System design Khalid Choukri

SL corpus design Els den Os

SL corpus collection Christoph Draxler

SL corpus representation Els den Os Part II:

SL lexica Dafydd Gibbon

Language models Hermann Ney

Physical characterisation and description Lars Knohl & Volker Kraft Part III:

Assessment methodologies and experimental design Peter Howell

Assessment of recognition systems David van Leeuwen & H. Steeneken

Assessment of speaker verification systems Frédéric Bimbot & Gérard Chollet

Assessment of synthesis systems Renée van Bezooijen & Vincent van Heuven

Assessment of interactive systems Norman Fraser Part III: Contributions from named Projects

Main new contributions and revisions Christoph Draxler, John Esling, John Wells

EAGLES I: Handbook production

• Hardback library edition + CD-ROM:
  • Publication date: late Spring 1997
  • CD-ROM hypertext: entire text + browser software
  • Price: about $180
• Four volume paperback edition:
  • Publication date: late Autumn 1997
  • Volumes contain (modified) user guide, a Handbook Part
  • Price: about $25 per volume
• Web edition (identical to CD-ROM)
• Price minimisation:
  • No royalties
  • Mixed calculation on both main editions
• Delay factors:
  • Numerous feedback cycles, size, heterogeneity of input
  • Page correlation between editions
  • Hypertext production problems: scale (first version about 2800 files), IPA problem (individual character gifs)
  • But fast (3 month) book production schedules.

`A good product should have a good package.'

EAGLES II: Responsibles and tasks

Dafydd Gibbon U Bielefeld Partner, WP5/6

Roger Moore DERA, Malvern WP5/6

Geoffrey Leech U Lancaster WP4

Richard Winski Vocalis, Cambridge WP5/6

Tasks:

WP4: LUK Integrated resources for spoken and written language

WP5: UBI (DRA) Revision and extension of Handbook

WP5.1: (UBI) DRA Design issues, content co-ordination

WP5.2: UBI VOC Chapter revisions and extensions of Handbook

WP5.3: UBI (DRA) Expert evaluation wrt starting level

WP5.4: UBI Synthesis report

WP6: UBI (DRA) Integration/editing

WP6.1: UBI Technical editing of print edition

WP6.2: UBI Technical editing of hypertext edition

WP6.3: UBI Synthesis report

EAGLES II: Planned Supplement

• Chapters
  1. Dialogue (integration with WP4) -- Geoffrey Leech
  2. Audio-visual/multimodal integration -- Christian Benoıt, Alex Waibel
  3. COTS (Commercial Off-The-Shelf products) -- Herman Steeneken
  4. Terminology -- Dafydd Gibbon
  5. Reference materials -- Lou Boves, Christoph Draxler
• Expert consultation meetings (next: January 1998)
• Two-phase authoring: I Spring 1998, II Summer 1998.
• Preliminary review via Commission by other EU projects, Spring 1998 (formal review policy still to be communicated).
• Preliminary version of Terminology chapter due as interim deliverable at end of 1997
• Output of WP 4, 5, 6: joint volume on Standards and Resources for Spoken Language Systems, comparable with Phase I paperbacks, draft output about September 1998