Word segmentation
(with Sharon Goldwater and others)
Before infants begin to learn to talk or to understand spoken language,
they must first identify individual words from fluent speech.
To
accomplish this challenging task, they use information from a variety
of sources, including the phonotactics and stress pattern of their
language and distributional information about predictability between
syllables or phones in sentences. I use computational models
of
segmentation and artificial language experiments with adults to map the
mechanisms by which humans accomplish the task of segmentation.
Word learning (with Noah Goodman and Josh Tenenbaum)
How do children learn
the words of their native language? One account posits that
infants observe regularities in the contexts of use for different words
across situations (Siskind, 1996), allowing for the learning of
consistent associations. We are attempting to use
computational
models both to characterize this process of cross-situational
observation and to model other experimental results in the domain of
early word learning.
Eye-tracking
and infant
perception
(with Ed Vul and Scott Johnson)
How do very young infants see the world? Do the same parts of
it
draw their attention as draw ours? To investigate
this
question, we tracked the eyes of a range of infants from 3- to 12
months old and have used a variety of computational tools to
characterize the patterns of fixation we observed.
Language, culture, and cognition
(with Daniel Everett, Ev Fedorenko,
& Ted Gibson)
What is the relationship between language and thought? We have studied
the cognition of the Pirahã, a tribe of indigenous people
living
in Amazonas, Brazil, with the goal of investigating some of the
surprising features of their language. For instance,
Pirahã is the first documented language which contains no
words
for numbers. In addition, we are investigating the claim (first
made in Everett, 2005) that Pirahã has no recursive syntactic
structure.
