Speaker: Roni Katzir (Tel-Aviv University & MIT)
Title: Structure and learning of quantificational determiners
Date and time: Wednesday, February 15, 1-2pm
Location: 32-D461
Abstract:
Acquiring semantic denotations — even the entry for a single, well-exemplified, low-type element — presents the child with a difficult inductive challenge. I start by illustrating this challenge using the notion of learning known as identification in the limit, before switching to a less complete notion of learning, compression-based learning, which offers a more constructive way to approach the inductive challenge. Focusing on the representation and learning of quantificational determiners, I show how compression-based learning maps representational choices — e.g., basic determiners and their combinations, in an intensional variant of Keenan & Stavi 1986 (following last week’s discussion), or semantic automata, as in van Benthem 1986 — onto learners. This mapping, in turn, makes empirical predictions that can help us choose between competing architectures.