Abstract Stephen McGregor 12 November 2018
TITLE: Are Computers Really Suited for Processing Natural Language? Recent Experiments in Grounded Language Learning SPEAKER: Stephen McGregor, Laboratoire Lattice, École normale supérieure/CNRS
ABSTRACT: In this presentation, I'll explore questions of how language is grounded in the course of an agent's encounter with the world, and present some early results from computational models targeting a new dataset of simulated human-robot interactions motivated by these questions. These experiments are designed to shed light on the way that conceptually rich, contextually sensitive representations might come about in the course of a language learning agent's interaction with an environment under the tutelage of informed members of a language community. I'll discuss the way in which these experiments might raise questions about the necessary and sufficient architecture of a language using agent, potentially providing a platform for examining the emergence, on developmental and evolutionary timescales, of intentional and goal-directed behaviour.
BIO: Stephen McGregor is currently a Research Associate working with Laboratoire Lattice at the École normale supérieure, as a member of the CHIST-ERA project Atlantis. He studied for his PhD at Queen Mary from 2013 to 2017 as a member of the Computational Creativity Lab and the Computational Linguistics Lab, co-supervised by Geraint Wiggins and Matthew Purver. In addition to his research on topics surrounding the emergence of contextually productive lexical semantics in a dynamic environment, he is involved in ongoing work relating to digital humanities as well as the philosophical foundations, applications, and implications of computational linguistics.