Abstract
In NLG, there has so far been little emphasis on building domain models; but in order to arrive at the lexical variation in verbalizations that we have set out to achieve, the decisions for domain modelling need to be made carefully. Thus, Chapter 4 develops the domain model for the generator: the taxonomy of concepts and relations that, in later chapters, the generation system will be based on.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Notes
In this book, we will not be concerned with simulation, though.
Both concepts and instances are pre-linguistic entities and thus appear in SMALLCAPS. To distinguish them, we will adopt the convention of forming instance names by adding numbers to the name of the concept they are instantiating. For example, DOGl and DOG2 would stand for (distinct) instances of DOG.
Dale [1992] provided an extensive analysis of kinds of objects (in the domain of cooking recipes) and their role in language generation. Our work, focusing instead on event verbalization, can be seen as complementary.
In this section, it will turn out somewhat difficult not to be confused by the various terminologies for situations and their subtypes. Before our own categories are defined, we will use occurrences as a generic, theory-neutral term referring to the things that can be “going on” in the world—exactly those that need to be classified here.
Parsons [1990], however, reminds us that work on verb classification had indeed started several centuries before Christ, and that in our century, amongst others, Russell and Ryle have investigated some of the distinctions later elaborated by Vendler.
Throughout this section, a large number of terms yearn to be set in italics and SMALL-CAPS simultaneously, because they are fixed terms used by other writers on the one hand, and refer to pre-verbal entities on the other. We opt for the solution of using italics when discussing the work of others, and later returning to SMALLCAPS when the actual categories of our own system are introduced.
For an approach to automatically understanding instructions, also using representations in a DL, see Di Eugenio [1998].
In fact, LOOM offers some operators for reasoning with such intervals; we could, for instance, define a subtype like COOL-TEMPERATURE-STATE, where the range might be (: through 0 10), and whenever some TEMPERATURE-STATE is created, the classifier would assign either the specific or the more general concept to it, depending on the particular value.
This is a shortcut; a more sophisticated model would treat only the TANK-OPENING as a CONNECT-PART, but not the entire TANK.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1999 Springer Science+Business Media New York
About this chapter
Cite this chapter
Stede, M. (1999). Modelling the Domain. In: Lexical Semantics and Knowledge Representation in Multilingual Text Generation. The Springer International Series in Engineering and Computer Science, vol 492. Springer, Boston, MA. https://doi.org/10.1007/978-1-4615-5179-9_4
Download citation
DOI: https://doi.org/10.1007/978-1-4615-5179-9_4
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4613-7359-9
Online ISBN: 978-1-4615-5179-9
eBook Packages: Springer Book Archive