Abstract
The representations of situations in the domain model, as introduced in the last chapter, are quite distant from specific natural language sentences, and trying to map these structures directly to linguistic output would not be a promising endeavor. Very many decisions need to be made, and, particularly in a multilingual environment, a lot of work would be duplicated if language-specific modules were in charge of the complete mapping, because many decisions will be identical for all the target languages. This chapter thus argues for a division between a language-neutral level of situation specification and an intermediate, language-specific level of semantic sentence representation. By drawing upon language-specific lexical resources, a single language-neutral algorithm can produce the semantic representations for both target languages; specifications on this level can then be processed by surface generators and converted to individual sentences.
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
When embedding the system in some application involving user models and other pragmatic information, the high-level communicative goals (or ‘rhetorical goals’ in the terminology used by Hovy [1988a]) are to be put into correspondence with the semantic goals; but that is beyond our project here.
The PENMAN system nowadays has been joined with the German ‘Komet’ grammar [Teich 1992] into the multilingual generation system KPML (‘Komet Penman Multilingual’) [Bateman 1997], which serves as the official “follow-up” to PENMAN.
More accurately, realizations with different meaning stem from different UM types. Henschel [1993] points out that “disjoint concepts in the UM do not necessarily correspond to disjoint sets of surface sentences—only to disjoint semantic perspectives on them. The interface between the UM and the grammar should be written in such a way that it is possible in some cases to generate the same sentence from different semantic input.”
In systemic-functional linguistics, which is the theoretical basis of PENMAN, there is in theory no separation between lexicon and grammar; both are intertwined in the network of choice points (’systems’), the lexicogrammar.
The distinction between participants and circumstances is made in one way or another in any linguistic theory, where the realizations as surface constituents are, for example, called complements and adjuncts. The former are seen as being subcategorized for by the verb, whereas the latter are not. We will discuss these notions in Section 6.3.
Further arguments are given by Stede and Grote [1995].
The more recent development of a ‘Generalized Upper Model’ [Bateman et al. 1994] appears to take a new stand on the issue of language-specificity.
See also Henschel [1993], who describes efforts on merging an English and a German UM.
Which keywords correspond to participants and which to circumstances depends on the process type. We will say more about the distinction in the next chapter.
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). Levels of Representation: Sitspec and Semspec. 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_5
Download citation
DOI: https://doi.org/10.1007/978-1-4615-5179-9_5
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4613-7359-9
Online ISBN: 978-1-4615-5179-9
eBook Packages: Springer Book Archive