Thèse en Traitement Automatique du Langage Naturel/Intelligence Ambiante

Funded PHD thesis in Computational Linguistics/Ambient Intelligence Grenoble, France


Keywords: Natural language understanding, speech processing, decision making, smart home

The Laboratoire d’Informatique de Grenoble (LIG) of the University Grenoble Alpes, Grenoble, France invites applications for a PhD position in Computational Linguistics for Ambient Intelligence.

The position starts in September 2017 and ends in July 2020 and is proposed in the context of the national project Vocadom ( whose aim is to build technologies that make natural hand-free speech interaction with a home automation system possible from anywhere in the home even in adverse conditions [Vacher2015].

The aim of the PhD will be to build a new generation of situated spoken human machine interaction where uttered sentences by a human are understood within the context of the interaction in the home. The targeted application is a distant speech hand free and ubiquitous voice user interface to make the home automation system react to voice commands [Chahuara2017]. The system should be able to process possibly erroneous outputs from an ASR system (Automatic Speech Recognition) to extract meaning related to a voice command and to decide about which command to execute or to send a relevant feed-back to the user. The challenge will be to constantly adapt the system to new lexical phrases (no a priori grammar), new situations (e.g., unseen user, context) and change in the house (e.g., new device, device out of order). In this work, we propose to extend classical S/NLU (Natural Language Understanding) approaches by including non-linguistic contextual information in the NLU process to tackle the ambiguity and borrow zero-shot learning techniques [Ferreira2015] to extend the lexical space on-line. Reinforcement learning is targeted to adapt the models to the user(s) all along the use of the system [Mnih2015].

The candidate will be strongly encouraged to publish their progress to the main events of the field (ACL, Interspeech, Ubicomp).

The PhD candidate will also be involved in experiments including real smart-home and real users (elderly people and people with visual impairment) [Vacher2015].


[Mnih2015] Mnih, Kavukcuoglu et al.  Human-level control through deep reinforcement learning. Nature 518, 529–533
[Chahuara2017] Chahuara, F. Portet, M. Vacher Context-aware decision making under uncertainty for voice-based control of smart home Expert Systems with Applications, Elsevier, 2017, 75, pp.63-79.
[Ferreira2015] E Ferreira, B Jabaian, F Lefevre Online adaptative zero-shot learning spoken language understanding using word-embedding Acoustics, Speech and Signal Processing (ICASSP), 2015
[Vacher2015] M. Vacher, S. Caffiau, F. Portet, B. Meillon, C. Roux, E. Elias, B. Lecouteux, P. Chahuara. Evaluation of a context-aware voice interface for Ambient Assisted Living: qualitative user study vs. quantitative system evaluation. ACM – Transactions on Speech and Language Processing, Association for Computing Machinery, 2015, pp.5:1-5:36.

– Master’s degree in Computational Linguistics or Artificial Intelligence (Computer Science can also be considered)
– Solid programming skills,
– Good background in machine learning,
– Excellent English communication and writing skills,
– Good command of French (mandatory),
– Experience in experimentation involving human participants would be a plus
– Experience in dialogue systems would be a plus plus

Applications should include:
– Cover letter in French outlining interest in the position
– Names of two referees
– Curriculum Vitae (CV) (with publications if applicable)
– Copy of the university marks (grade list)

University of Grenoble Alpes is situated in a high-tech city located at the heart of the Alps, in outstanding scientific and natural surroundings. It is 3h by train from Paris ; 2h from Geneva and is less than 1h from Lyon international airport.

Research Group Website :

For further questions, please send an email to