Kevin Denamganaï

Kevin Denamganaï

Natural Language Interaction in Video Games

We make sense of the world around us through words, independently of them being spoken, written, or even sung. Yet, so far, we rarely play video games through our spoken words. We might acquire pieces of information from the game through one-way scripted

natural language interactions, but this interaction is never improvised, never coming from the player, and it is certainly not central to the gaming experience.

Thus, I aim to research about the introduction of two-way vivid natural language interactions in video games through, for instance, a novel way of interacting with any pre-existing video game : the controllers would no longer be handed off to the player but rather to an artificial agent. While both look at the video game, only the agent can influence the original video-gaming experience. The role of the player is to communicate with the agent in order to teach it how to beat the game, through co-operative dialogues and a form of mutual understanding.

This work could bring about a novel kind of gaming experience that puts the emphasis on communication/teaching skills, and it could yield advances in the field of human-computer/robot interfaces designed to learn, at a real-time scale, through human feedback.

After graduating as an Engineer from the Ecole Nationale Supérieure de l'Electronique et de ses Applications (ENSEA), France, with two double-degree diplomas, a MEng in Electrical Engineering and Information Science from the Osaka Prefecture University (OPU), Japan, and a MRes in Artificial Intelligence and Robotics from the Université de Cergy-Pontoise (UCP), France, I have spent a year accumulating experience as a Robotics & Machine Learning freelancer. I am now looking forward to put those skills at use in the IGGI PhD program, that, among other things, gives me the opportunity to reunite with video games. Indeed, it was thanks to a keen interest towards video game creation that I started learning programming at 12.

My research interests are about everything psychology, neuroscience, AI, (deep) reinforcement learning, robotics, and more recently natural language understanding as well as human-computer interfaces, challenging the question what are the necessary components of an artificial agent to be able to converse with human-beings in an engaging manner, given the assumption of a pre-defined goal that needs to be fulfilled, e.g. cooperatively clearing a level in any given video game.

Home institution: York

Supervisor: Dr Daniel Kudenko

Ready to apply?

Once you have identified your potential supervisor, we would encourage you to contact them to discuss your research proposal.

Learn More