Title: A Crow's Beak is not Yellow - Investigations on Cognitively Salient Concept Properties

Gerhard Kremer

Abstract:

Concept descriptions that are collected from participants in experiments (e.g.,
rabbits: "they are mammals, they have long ears, they jump,...") are widely 
used by the cognitive science community:

These feature sets (comprising concept properties of different types (e.g., 
category, parts, behaviour,...) are assumed to indicate how concepts are 
represented in the human mind, and thus they serve in simulations of cognitive
tasks and experimental design.

In the area of computational linguistics, semantic models based on these human-
generated properties have been shown to be a good complement or alternative to
traditional semantic models based on corpus collocates.

In my PhD project, I (i) empirically investigated which features participants
systematically produced in a concept description experiment (comparing results
of the two target languages German and Italian) and (ii) conducted a case study
on automatically extracting such cognitively salient properties for a given 
concept from text corpora. The case study focused on (adjectival modifiers of)
composite expressions for constitutive part properties (e.g., crow: has a
_black_ beak).