Wednesday, February 10th, 2010
Knowing versus naming: Similarity and the linguistic categorization of artifacts – Malt et al 1999
How associated are similarity and linguistic categorization?
Previous work suggest some parallels but also some disassociations.
Previous work found that we judge similarity by perceptual features but name categories using more functional features (Rips 1989)
Chaining(Lakoff 1987): Similarity between a named object and it’s nearest neighbor in similarity space may determine linguistic categorization. This allows objects that are different from category exemplars to be members of the category if there is a salient chain of other category members that are more similar.
Convention: Certain objects may retain linguistic categories that defy perceptual similarity with other categories based on linguistic history. Could arise from chains that are no longer apparent.
Pre-emption: A linguistic category may be used to pre empt ambiguity with another salient category even if the other category is in the same region of similarity space. (Clark)
“Although perceived similarity and naming show a positive relation, this relation is far from perfect and factors other than similarity must contribute to naming choices.”
What if linguistic naming tasks also took into account dissimilarity along with similarity?
Wednesday, February 10th, 2010
Towards a frame based lexicon: The case of RISK – Charles J. Fillmore and B. T. Atkins (1992)
Introduces the notion of frame semantics. A word’s meaning can be understood only with reference to a structured background experience, beliefs, or practices, constituting a conceptual prerequisite for understanding the meaning. The meaning of a word can only be understood by first understanding background frames that motivate the concept that the word encodes. Words are not related directly but by their links to common background frames. Framenet attempts to document the entire set of frames a given lexical item can relate to and how the frame elements are syntactically realized. Seems to differ from the notion of scripts in that it is non episodic.
Wednesday, February 10th, 2010
Scripts, Plans, Goals, and Understanding: An Inquiry into Human Knowledge Structures – Schank & Abelson 1977
Representations of meaning must make all implicit information explicit. Uses decomposition into semantic primitives. Noun semantics can be reconstructed from episodic appearances. Rejects semantic memory systems in favor of episodic representations. Uses meaning postulates between semantic primitives to capture causal inference. Highly connected states and events in the conceptual dependency causal graph can be considered impotent. Understanding is a process by which people match what they see and hear to pre-stored groupings of actions that they have already experienced.