ISCA - International Speech
Communication Association

ISCApad Archive  »  2011  »  ISCApad #154  »  Jobs  »  (2011-03-01) Post-doctoral position in Automatic Speech Recognition at IRISA, France.

ISCApad #154

Friday, April 22, 2011 by Chris Wellekens

6-18 (2011-03-01) Post-doctoral position in Automatic Speech Recognition at IRISA, France.
Post-doctoral position in Automatic Speech Recognition at IRISA, France.

In the framework of the project 'Automatic System Harnessing' (ASH),
funded by the French National Research Agency, IRISA seeks to hire a
postdoctoral researcher in the field of automatic speech
recognition. Landmark-driven paradigms for large vocabulary speech
recognition will be the main topic of interest [1,2,3], with a focus
on developing bidirectional interaction mechanisms between landmark
detection and speech transcription. 

The ASH project is a national project aiming at developing
collaborative approaches to speech recognition by tight coupling of
several ASR systems or by coupling an ASR system with landmark
detectors. The idea of coupling (harnessing) systems originates from
the driven decoding principle as explained in [4]. The project is
leaded by the Laboratoire d'Informatique de l'Université du Mans
(LIUM) with the Laboratoire d'Informatique d'Avignon (LIA) as partner
in addition to IRISA.

The postdoc will be in charge of designing and implementing
interaction mechanisms between the IRISA's ASR system and that of the
other partners, during the decoding stage. Similar mechanims will be
investigated for joint ASR and landmark detection in Irisa's ASR
system. These research activities will be incoporated in fit in the
framework of the upcoming national benchmark ETAPE/ESTER 3.

Candidate should be a recent PhD in computer science or electrical
engineering with strong experience in large vocabulary speech
recognition (preferably search or acoustic modeling) and/or audio
signal processing and classification. Candidates should have good
practical and theoretical knowledge in pattern recognition. We are
primarily interested in statistical approaches to pattern recognition
but skills in discriminative techniques, in particular neural
networks, are welcome. As we are mostly dealing with French data,
knowledge of the French language is a plus but not required.

IRISA is a French public laboratory located in Rennes (France),
affiliated to CNRS, the French national Center for Scientific
Research. The candidate will join the leading multimedia group TEXMEX
at IRISA ( and develop his research in strong
collaboration with the speech and audio processing group METISS. He
will be part of the ASR&NLP workgroup which gathers leading scientists
in speech recognition, natural language processing and information
retrieval. Work will be carried out in close collaboration with a PhD
student on landmark-driven speech recognition and in interaction with
ASH partners.

The position, to be fulfilled as soon as possible, is opened for one
year, possibly with a 6 months extension. Gross salary is 2500 euros
per month (2040 net pay) and includes social benefits.

Please contact Guillaume Gravier ( for more details or
send an extended resume for application.

[1] M. Hasegawa-Johnson, S. Borys and K. Chen, Experiments in
Landmark-Based Speech Recognition. Sound to Sense: Workshop in Honor
of Kenneth N. Stevens, June, 2004.

[2] G. Gravier and D. Moraru. Towards phonetically-driven hidden
Markov models: Can we incorporate phonetic landmarks in HMM-based ASR? 
In Proc. ISCA Tutorial and Research Workshop on Non Linear Speech
Processing (NOLISP), 2007.

[3] S. Sainath. Island driven search using broad phonetic
classes. IEEE Worskhop on Automatic Speech Recognition and
Understanding, 2009.

[4] B. Lecouteux, G. Linarès, Y. Estève, and G. Gravier. Generalized
driven decoding for speech recognition system combination. In IEEE
Intl. Conf. on Acoustics, Speech and Signal Processing, 2008.

Back  Top

 Organisation  Events   Membership   Help 
 > Board  > Interspeech  > Join - renew  > Sitemap
 > Legal documents  > Workshops  > Membership directory  > Contact
 > Logos      > FAQ
       > Privacy policy

© Copyright 2024 - ISCA International Speech Communication Association - All right reserved.

Powered by ISCA