This is a past event
What are the decision making processes by which humans choose words to achieve goals in their environment? Can we replicate them in a computer system? These are the core questions of grounded dialogue modelling, the study of interactive language use in context. We will start by looking at a few examples of grounded, goal-oriented dialogues. Then we will present a formal framework that incorporates fundamental notions of linguistic interaction— such as goals, costs, utility, and context — into an information theoretic model of communication. Using visually grounded referential dialogues as a case study, we will show how these notions can be modelled with state-of-the-art computational techniques, and how they explain human strategies of language production in terms of communicative efficiency and effectiveness. We will conclude by discussing the impact of our framework on language technologies, focusing in particular on the evaluation and development of natural language generation systems.
- Mario Giulianelli
- Meston 3 and MS Teams
Please contact Ehud Reiter for more informatation (firstname.lastname@example.org)