Statistics Seminars: Uncertainty measures for vague linguistic information
12 May 2006 00:00 in CM107"This talk will give an overview of the label semantics framework for modelling uncertainty associated with vague description labels. In contrast to fuzzy logic and other multi-valued logic approaches,label semantics encodes the meaning of linguistic labels according to how they are used by a population of communicating agents to convey information. From this perspective, the focus is on the decision making process an intelligent agent must go through in order to identify which (if any) labels are appropriate to use in order to describe a particular object or value. Central to this approach is an epistemic stance, according to which individual agents assume the existence of a set of conventions governing appropriate label use, which should be adhered to during communications, but which are only partially known to them. This approach is similar to the `anti-representational' view of vague concepts proposed by Parikh,and also to the work of Kyburg and Williamson.
A formal calculus is proposed based on two inter-related measures quantifying an agents subjective belief as to the appropriateness of labels for a given instance. It shown that this calculus can never be truth functional, but can be functional in a weaker sense. Within this framework we then consider how agents might identify a particular appropriate expression to assert, and what information may be conveyed by such assertions."
CM107 @ 3pm
Contact firstname.lastname@example.org for more information