Inhalt des Dokuments
|Olivier Pietquin, Supélec Metz|
|Monday, May 6, 2013, TU Hochhaus, 20th floor, Auditorium 1|
|Host: Klaus-Peter Engelbrecht|
Spoken dialogue systems are speech-based human-machine interfaces most often designed to complete a specific task (book a flight ticket, find touristic information, museum guides, weather forecast by phone etc.). The design of such a system is a very challenging task since it should take into account task-specific, user-specific and system-specific constraints. It is thus hard to assess and generally has to be done again for every novel task. For these reasons, user simulation has been an active topic of research for decades. It aims at generating simulated interactions with the system so as to quantify its quality or to train it in the case of strategy learning systems. In this talk, we will focus on statistical user simulation and will focus on a novel paradigm, inverse reinforcement learning, that aims at imitate human users from collected data. This method is also expected to simulate the co-adaptation occurring when the user adapts to the system he’s interacting with.
Olivier Pietquin obtained an Electrical Engineering degree from the Faculty of Engineering, Mons (FPMs, Belgium) in June 1999 and a PhD degree in April 2004. In 2011, he received the "Habilitation à Diriger des Recherches" (French Tenure) from the University Paul Sabatier (Toulouse, France). Now he is a Professor at the Metz campus of the Ecole Superieure d'Electricite (Supelec, France), where he headed the "Information, Multimodality & Signal" (IMS) research group from 2006 to 2010 when the group joined the UMI 2958 (GeorgiaTech - CNRS). In 2012, he became head of the Machine Learning and Interactive Systems group (MaLIS). Since 2010, Olivier Pietquin sits at the IEEE Speech and Language Technical Committee and he is a Senior IEEE member since 2011.