|
OBJECTIVES
Objectives of the
research
a.
To
develop a connectionist model of spatial language which deals with both
geometric and extra-geometric components and which maps onto real
psycholinguistic data.
b.
To
build a semantic feature database based on empirical studies which map out
how important particular properties (e.g., size, function) are for
particular objects, and the relative extent to which use and comprehension
is based on these properties versus geometry in a visual scene.
WORKPLAN
Phase 1 (months 1-8): Design of
computation model + Preliminary experiments on object features/functions
Deliverables (see publication
page for full details):
-
Analysis
and design specification for the whole model + initial work on vision
processing module.
-
Preliminary
modelling: IJCNN2001 and WCNI2001 conference papers.
-
Experimental
results: 2 chapters in the Indiana workshop proceedings
Phase 2 (months 9-20): Modelling
of perceptual grounding and time dynamics + Experimental work on spatial
prepositions over, under, above, below and on time course processing
Deliverables:
-
Model
results: data on vision module (ICONIP03 conference paper) and Elman
network (ICCM03 conference paper)
-
Experimental
results: J. Child Lang., Cognitiva, 2 Chapters in Carlson & van der
Zee’s book
Phase 3 (months 21-24):
Modelling spatial terms and time prediction data + Experimental testing of
model predictions
Deliverables:
-
Model
+ experiment overview: J. Memory and Language). (submitted) paper + other
papers in preparation (e.g. Cognitive Psychology)
-
Paper
to be presented at Neural Computational and Psychology Workshop
|
