EECS researcher wins over £1million of funding for research project
Congratulations to Massimo Poesio, Professor of Computational Linguistics, who has been awarded £1,016,382 through an EPSRC grant for a new project, entitled ‘Annotating Reference and Coreference in Dialogue Using Conversational Agents in Games’.
The grant, awarded by the Engineering and Physical Sciences Research Council (EPSRC), the main funding body for engineering and physical sciences research in the UK, will help fund the project for the next three years.
The project focuses on conversational agents, also known as chatbots, which are computer programs designed to simulate human text or verbal conversations. The technology behind conversational agents has made huge steps forward in recent years thanks to the adoption of deep learning approaches, but a number of issues remain.
Traditionally, training conversational agents requires a lot of data, so this project will adopt a human-in-the-loop approach, where conversational agents will learn from interactions with humans. The project is interested in improving the conversational agents use of two aspects of language interpretation: anaphoric interpretation and reference. In a new approach, the conversational agents will be developed and embedded into popular videogames like Minecraft, to interact with the players.
The project combines expert knowledge from both NLP (Natural Language Processing) and Games and AI areas of research. Dr Julian Hough and Professor Matthew Purver will utilise their expert knowledge in NLP, including their work developing dialogue systems and anaphoric interpretation, alongside Dr Diego Perez-Liebana from the Games and AI research group, who has an extensive track record in embedding agents into videogames.
In addition, other collaborations to the project include the University of Essex, who have been developing Games-With-A-Purpose (GWAPs) for almost fifteen years, Professor Julia Hockenmaier from the University of Illinois at Urbana, who pioneers the use of Minecraft for collecting dialogue data, and Dr Jack Urbanek from Facebook AI, who has been developing a light game platform for collecting data via games.
The project, if successful, aims to push forward the state of the art of research on anaphora in dialogue, and on dialogue in general, and also hopes to create conversational agents in games for future projects.