Spoken Language Interaction in Telecommunication at ENST/CNRS-LTCI

1
Spoken Language Interaction in Telecommunication
at ENST/CNRS-LTCI
Gérard CHOLLET, Richard CROCE, Dijana
PETROVSKA-DELACRETAZ, Marc SIGELLE, Pascal
VAILLANT, François YVON (chollet,croce,petrovsk,s
igelle,vaillant)_at_tsi.enst.fryvon_at_infres.enst.fr E
NST/CNRS-LTCI46 rue Barrault75634 PARIS cedex
13http//www.tsi.enst.fr/chollet
2
Outline

• What is ENST/CNRS-LTCI ?
• Research and application topics
• The SIROCCO project
• The EUREKA !2340 MAJORDOME project
• VoIP, VoiceXML, Human-Computer Interaction
• Perspectives

3
Our affiliations
ENST Ecole Nationale Supérieure des
Télécommunicationshttp//www.enst.fr CNRS
Centre National de la Recherche
Scientifiquehttp//www.cnrs.fr LTCI
Laboratoire de Traitement et Communication de
lInformation
4
What is ENST?Ecole Nationale de
Télécommunications

• classed among the
• Grandes Ecoles d'Ingénieurs.
• 250 state certified engineers
• each year .
• part of Groupement des Ecoles
• de Télécommunications

5
GET Groupement des Ecoles de
Télécommunication

• ENST
• ENST-Bretagne in Brest
• Institut National des Télécommunicationsin Évry
• Eurecom in Sophia-Antipolis
• ENIC (Ecole Nouvelle dIngénieurs en Télécoms)
  in Lille
• Institut des Applications Avancées de lInternet
  in Marseille

6
Academic departments within ENST

• COMELEC Communications, Electronic, VLSI,
• INFRES Computer Science, Networking, NLP,
• TSI Signal and Image Processing, Speech,
• EGSH Economy, Management, Social Sciences,

7
TSI Department Signal and Image Processing

• "Image Processing and Understanding"
• "Statistical Signal Processing Applied to
  Communications"
• "Perception, Learning and Modelling"  
• Very Low Bit Rate Speech Coding
• Speech Recognition, Speaker Verification
• "Coding"  
• Speech and Sound compression
• "Audio, Acoustics and Waves"  
• acoustical antennas, audio protheses

8
SIROCCO project Unlimited Vocabulary Speech
Recognition
INRIA (IRISA et LORIA), LIA, IRIT,
ENST-LTCI http//www.irisa.fr/sirocco/
9
SIROCCO

• Unlimited vocabulary speech recognition system
• French lexicon (MathLex) with 64kwords (AUF task)
• Feature extraction with Spro (G. Gravier)
• Context-dependent HMM phone models
• Word pronunciation graph
• Uses CMU-Toolkit for Language modeling
• Beam search for word hypothesis
• Rescoring of word hypothesis by A

10
EDF
Holistique
MAJORDOME Unified Messaging System Eureka
Projet no 2340
D. Bahu-Leyser, G. Chollet, R. Croce, K. Hallouli
, J. Kharroubi, D. Kofman, L. Likforman, E.
Matta-Sanchez, D. Petrovska, M. Sigelle, P.
Vaillant, F. Yvon
11
Majordomes Functionalities
12
Overview of Majordome

• Background tasks (server-side only)
• sorting and filtering messages from different
  sources (E-mail, voice, fax, SMS,)
• extracting relevant information for reporting to
  user (names of senders, subject,).
• Dialogue with the user over phone or Web.
• The system presents the state of the mailbox, the
  type of messages, their sender, subject, and may
  sum them up or read them on request
• The users access their mailbox, addressbook, time
  schedule, or URIs (Web addresses).

13
Voice technology in Majordome

• Server side background tasks
• continuous speech recognition applied to voice
  messages upon reception
• Detection of sender name and subject
• User interaction
• Speakers identification
• Speech recognition (receiving users commands
  through voice interaction)
• Text-to-speech synthesis (reading text summaries,
  E-mails or faxes)

14
Voice Over IP Platform
Network 192.168.222.0/11
Network192.168.223.0/11
Visioconference
15
Majordome partners
16
Majordome / NetCentrex project
PABX /Gateway ENST -Call Control
Server -Application Server
Calling person
Usual
NetCentrex
IP-VR NetCentrex Recorder Machine
No response
Usual user called
Vocal E-mail
17
Majordome / NetCentrex project
PABX /Gateway ENST -Call Control
Server -Application Server
Calling person
Usual
NetCentrex
Voice Interactive call
No response
MAJORDOME
IP-VR NetCentrex

• Speaker verification
• Dialogue
• Vocal e-mail
• Routing
• Updating the agenda
• Automatic summary

Usual user called
18
A framework A L I S P
A utomatic L anguage I ndependent S peech P
rocessing
with applications in Speech Coding, Synthesis,
Recognition, Speaker Verification and Language
Identification
19
Perspectives

• The application context of the Majordome project
  could be of interest to COST-278.
• The Majordome/NetCentrex platform could be made
  available to interested partners.
• HTK, ISIP and SIROCCO softwares are available as
  freeware. One of them will be used on the
  NetCentrex platform.