Post‐Doctoral : Multilingual Speech Translation Under Resource Constraints

Universities and Institutes of France
December 14, 2022
Offerd Salary:€3500 gross
Working address:N/A
Contract Type:Temporary
Working Time:Full time
Working type:N/A
Job Ref.:N/A
  • Organisation/Company: Université du Mans
  • Research Field: Computer science
  • Researcher Profile: First Stage Researcher (R1)
  • Application Deadline: 14/12/2022 18:00 - Europe/Brussels
  • Location: France › Le Mans
  • Type Of Contract: Temporary
  • Job Status: Full-time
  • Hours Per Week: 35
  • Offer Starting Date: 03/01/2023
  • This postdoctoral work is about multilingual speech translation under low- resources contraint. It is funded by a DGA project with the SYSTRAN company.

    Assigned activities and expected results :

    The objectives of this postdoc are multiple. Recent approaches to learning self-supervised representations such as Wav2Vec-2.0 Baevski et al., 2020 and many others Chung and Glass, 2020, Khurana et al., 2020, Conneau et al., 2020, Hsu et al., 2021, Babu et al., 2021, Chen et al., 2021, Chung et al., 2021, Bapna et al., 2022 provide state-of-the-art results on several spoken language tasks. These are large transformer-based models, trained on huge amounts of unlabelled audio data, and then refined on task-specific data. These models are mainly concerned with the representation at the acoustic level of the signal, i.e. for short segments of speech with a duration of 10 to 20 milliseconds. They do not seek to capture semantic information, which is necessary for tasks such as multilingual translation. One of the objectives of this work is to be able to encode both semantic and phonetic informations, in a multilingual way, in order to be able to carry out the task of speech translation. A first model has been proposed (SAMU-XLSR Khurana et al. 2022) and can be used as a basis for this work. The idea behind the use of a single model is to be able to benefit from a robust model, which will be able to work in a scenario where few annotated resources is avalaible for a given language. The postdoc will participate in evaluation campaigns in the field, and his work will be integrated into a demonstrator.

    To apply, you must consult the job description available on this link :

    https: // www.


    The postdoc will earn €3500 gross monthly.

    Offer starting date:January 03, 2023

    Contract length : 1 year with a possible extension

    Offer ending date:January 02, 2024

    The work will be done in the LIUM laboratory. The postdoc will have access to a computation facilities

    Web site for additional job details

    https: // www. recrutons...

    Offer Requirements
  • Communication sciences: PhD or equivalent

    Contact Information
  • Organisation/Company: Université du Mans
  • Department: LIUM
  • Organisation Type: Public Research Institution
  • Website: https: // https: // www. univ-
  • E-Mail:
  • Country: France
  • City: Le Mans
  • Street: Av Olivier Messiaen
  • Phone: 02 43 83 38 30
  • From this employer

    Recent blogs

    Recent news