Discourse pegs: a computational analysis of context-dependent referring expressions
Susann LuperFoy · University Microfilms International eBooks · 1992
A three-tiered representation of natural language discourse is proposed, which borrows from Landman's notions of pegs in a partial information state. Based on the three-tiered framework, a taxonomy of potentially context-dependent referring expressions is defined which includes definite pronouns, indefinite pronouns (one-anaphora), definite and indefinite noun phrases and proper names. Two parameters that classify uses of noun phrases into this taxonomy are (1) whether the context-dependent referring expression is totally anaphoric or partially anaphoric and (2) whether the expression is sponsored by elements in the linguistic tier or elements in the discourse tier. A set of algorithms for updating the computational discourse representation and interpreting anaphoric noun phrases is defined. The distinction is made between the interpreter's belief system, which in an implemented system is the underlying knowledge base, and the representation of the information content of the discourse itself. This distinguishes understanding a discourse from believing a discourse.