Starting job date (desired): March 2020. ================================================================== ## Work description
###Project Summary
Automation and optimisation of *verbal interactions of a socially-competent robot*, guided by its *multimodal perceptions*
Facing a steady increase in the ageing population and the prevalence of chronic diseases, social robots are promising tools to include in the health care system. Yet extant assistive robots are not well suited to such context as their communication abilities cannot handle social spaces (several meters and group of persons) but rather face-to-face individual interactions in quiet environments. In order to overcome these limitations and eventually aiming at natural man-robot interaction, the objectives of the work will be multifold.
First and foremost we intend to leverage the rich information available with audio and visual flows of data coming from humans to extract verbal and non-verbal features. These features will be used to enhance the robot's decision-making ability such that it can smoothly take speech turns and switch from interaction with a group of people to face-to-face dialogue and back. Secondly online and continual learning of the advanced system will be investigated.
Outcomes of the project will be implemented onto a commercially available social robot (most likely a Pepper) and validated with several in-situ use cases. A large-scale data collection will complement in-situ tests to fuel further researches. Essential competencies to address our overall objectives lie in dialogue systems / NLP, yet knowledges in vision and robotics would also be necessary. And in any case good command of deep learning techniques and tools is mandatory (including reinforcement learning for dialogue strategy training).
### Requirements
- Master or PhD in Computer Science, Machine Learning, Computational Linguistics, Mathematics, Engineering or related fields - Expertise in NLP / Dialog systems. Strong knowledge of current NLP / Interactive / Speech techniques is expected. Previous experience with dialogue and interaction and/or vision data is a strong plus. - Knowledge in Vision and/or Robotics are plusses. ? Strong programming skills, Python/C++ programmer of DNN models (preferably with pytorch) - Expertise in Unix environments - Good spoken and written command of English is required. *French is optional.* - Good writing skills, as evidenced through publications at top venues (e.g., ACL, EMNLP, SigDial etc) is a plus, for post-doc.
## Place
Bordered by the left bank of the Rhône Avignon is one of the most beautiful city in Provence, for some time capital of Christendom in the Middle Ages. The important remains of a past rich in history give the city its unique atmosphere: dozens of churches and chapels, the ?Palais des Papes? (palace of the popes) the most important gothic palace in Europe), the Saint-Benezet brigde, called the « pont d?Avignon » of worldwide fame through its commemoration by the song, and the ramparts that still encircle the entire city, ten museums from then ancient times to contemporary art.
The 94,787 inhabitants of the city, about 12,000 live in the ancient town centre surrounded by its medieval ramparts. Avignon is not only the birthplace of the most prestigious festival of contemporary theatre, European Capital of Culture in 2000, but also the largest city and capital of the département of Vaucluse. The region offers a high quality of urban life at comparatively still modest costs. In addition to this, the region of Avignon also offers the opportunity to visit numerous monuments and natural beauty sites easily accessible in a very short time: Avignon is the ideal destination for visiting Provence.
The position carries no direct teaching load, but if desired, teaching BSc or MSc level courses is a possibility (paid extra hours), as is supervision of student dissertation projects.
Initial employment is 12 months, extension is possible. For engineer, shift to a PhD position is possible.
## Applications
No deadline: applications are possible until the position is filled.
* Statement of research interests that motivates your application * CV, including the list of publications if any * Scans of transcripts and academic degree certificates * MSc/PhD dissertation and/or any other writing samples * Coding samples or links to your contributions to public code repositories, if any * Names, affiliations, and contact details of up to three people who can provide reference letters for you
