It is well known that participants in dialog use metalanguage -- that is, they talk about the dialog itself, or specific words in the dialog: as in, "What do you mean by X?" or "When I say 'photoreceptors' I mean both rods and cones." What is not specifically known is how often, or in what specific contexts people tend to use metalanguage, nor if there are reliable linguistic markers which signal the use of metalanguage.

This straightforward project involves the annotation of dialog corpora for instances and kinds of metalanguage, and various studies to learn something about its frequency and context.

This project is part of a larger, ongoing effort to develop flexible, interactive, adaptable and error-tolerant natural language computer interfaces.