*Title: Machine Learning for Adaptive Language interaction with an intelligent dialog companion
*Keywords: Deep reinforcement learning, interactive machine learning, natural language processing, neural networks, intelligent dialog companion
*Context: The thesis takes place in the context of a collaboration between the Robot Cognition Laboratory (INSERM/CNRS), the innovation services department of Intrinsic Cloud Temple, and the Flowers laboratory (Inria).
*Background: Based on RCL experience in human-robot cooperation and learning, we have developed an intelligent agent that accompanies humans in their experience using web-based tools, learns from explicit demonstration, and allows the user to re-use and recompose learned plans to form more complex structured plans.
*Project: For the moment, the system learns from demonstration. The project will extend and study algorithms for interactive machine learning (using Deep RL) to include language instructions as one of the ways that the system can learn, that is, by using the invariant structure of the human verbal description to organize procedural representations. This will advance the state of the art in the integration of language into reinforcement learning systems.
* Location: Lyon and Nanterre, France (Inserm and Cloud Temple)
* Response modalities: Send (1) a letter of motivation stating your interest in the project, (2) a CV, and (3) two letters of recommendation to the three addresses for Jean-Michel Dussoux, Peter Ford Dominey, and Pierre-Yves Oudeyer listed in the call, with the subject [Intelligent Assistant PhD].
jmd@intrinsec.com;
pierre-yves.oudeyer@inria.fr;
peter.dominey@inserm.fr