This paper describes the use of File Change Semantics, equivalently Discourse Representation Theory, as the theoretical basis for the discourse component of spoken dialogue systems. The approach was first developed in support of a non-speech, mixed-modality interface in which allowable input included mouse gestures for deictic reference to screen objects interspersed with keyboard entry of English and command language atoms. In the spirit of its theoretical foundation, this computational discourse framework defines the semantic value of each utterance to be its context change potential, i.e., its affect on the pre-existing discourse representation. The frame-work introduces modified file cards called Discourse Pegs, to serve as loci for aggregation of all information available to a dialogue agent at run time. Pegs hold linguistic (surface form), discourse, and world knowledge information in three separate tiers, each of which is structured, updated, and accessed via its own set of operations. The discourse tier houses intentional and attentional structures posited by Grosz and Sidner, while the linguistic structure is carried on the surface form tier and follows Clark and Schaefer (1987). A strength of our discourse framework is its ability to represent incomplete or incorrect information which can be updated incrementally and non-monotonically as the discourse progresses. This makes it especially well suited to spoken language systems in which speech recognition, parse, and semantic analyses are often uncertain or ambiguous and in which multiple discourse processing tasks are required of a single system. This paper describes the computational discourse framework and its adaptation to support interpreting telephony.
Cite as: LuperFoy, S. (1995) Implementing file change semantics for spoken-language dialogue managers. Proc. ESCA Workshop on Spoken Dialogue Systems, 181-184
@inproceedings{luperfoy95_sds,
author={Susann LuperFoy},
title={{Implementing file change semantics for spoken-language dialogue managers}},
year=1995,
booktitle={Proc. ESCA Workshop on Spoken Dialogue Systems},
pages={181--184}
}