On the Place of Words In the Generation Process
Generate an AI Snapshot to get a quick, structured summary of this paper.
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
Preliminary research on “lexical choice” in generation: the relationships between words and the representations employed by the speaker for reasoning and modeling the situation is described.
Abstract
This paper describes preliminary research on "lexical choice" in generation: the relationships between words and the representations employed by the speaker for reasoning and modeling the situation. We hold that the bulk of the variance that we see at the surface level in language has its origins very deep in the conceptual system. Consequently, most of the burden of lexical choice must be taken on by this internal representation, and not by the generator proper as is customary today. This work is exploratory rather than comprehensive. It studies the design consequences of three examples of lexical choice, introducing the devices "lexical clusters" and "action chain" as part of the representational system that organizes the selection process. In the first example the choice follows directly from the speaker's categorial judgements. The next describes how the very same information can receive two substantially different lexical realizations depending on the speaker's attitude toward it. In the last a choice usually ascribed to the generator is reanalyzed as conceptual, leading to a simpler processing architecture. These studies lend support to the conclusion that the selection of key lexical items is the first step in generation, with the choice criteria taken almost exclusively from the conceptual model and intentional attitudes of the speaker.
