|
BACKGROUND
Functional and
Extra-Geometrical Factors in Spatial Language and Cognition
Expressions involving spatial prepositions in
English convey to a hearer where one object (located object) is located in relation
to another object (reference object). For example, in the apple is in the
bowl, the apple is understood to be located with reference to the bowl in
the region denoted by the preposition in. Understanding the meaning of such
terms is important as they are among the set of closed class terms which
are generally regarded as having the role of acting as organising structure
for further conceptual material (Talmy, 1983). Furthermore, from the
semantic point of view spatial prepositions have the virtue of relating in
some way to visual scenes being described, and therefore measurable
characteristics of the world (Regier, 1996). Hence, it should be possible
to offer more precise semantic definitions of these as opposed to many
other expressions because the definitions can be grounded in perceptual
representations.
Most approaches to spatial prepositions have
assumed that they only require coarse grained properties of the objects
involved as constraints on their use (e.g., Herskovits, 1986; Landau and
Jackendoff, 1993). Computational models too have made the same assumption,
and have focussed on mapping individual prepositions onto geometric
computations in the scene being described (e.g., Logan & Sadler, 1996;
Regier, 1996; Regier & Carlson, 2001; Gapp, 1994). Yet there is now
much evidence (see Coventry & Garrod, 2004, for a comprehensive review)
that “what” objects are influences how one talks about “where” they are.
For example, Coventry, Prat-Sala and Richards (2001) found that whether an
object was shown to be fulfilling its protecting function influenced the
appropriateness of over, under, above and below to describe the relative
positions of the umbrella and man shown in Figure 1. These prepositions
were most appropriate to describe scenes where the umbrella was shown to
protect the man from getting wet and least appropriate when the man was
shown to get wet.
Figure 1.
Sample scenes used Coventry, Prat-Sala and Richards (2001)
The main aims of the grant were to develop a connectionist
model of spatial language which deals with both geometric and
“extra-geometric” constraints and maps onto real psycholinguistic data, and
to establish empirically the relative extent to which these constraints
influence the comprehension of a range of spatial prepositions. In the
early stages of the grant it became apparent that we needed to address the
theoretical goal of bridging the gap between theories of meaning which
capture meaning in terms of symbol-symbol relations (e.g., Landauer & Dumais,
1997) versus those which “ground” language directly in perceptual
representation (e.g., Regier, 1996). We wanted to generate a model which
encapsulated symbol-symbol processing and visuo-symbol processing, thus
grounding meaning in both perceptual representation and symbol-symbol
relations.
|
