John A. Bateman, Renate Henschel, Fabio Rinaldi
The Generalized Upper Model 2.0 is a further evolution of the Generalized (English,German,Italian) Upper Model first described in Bateman et al. (1994) and Bateman et al. (1995), which was in turn an outgrowth of the Penman Upper Model (Bateman et al., 1990) and the Merged Upper Model for English and German (Henschel, 1993; Henschel and Bateman, 1994). We also now incorporate aspects of the theoretical basis articulated at length by (Halliday and Matthiessen, to appear), thus bringing the Generalized Upper Model more closely in line with the Ideation Base.
The Generalized Upper Model is presently used in the Multilingual Generation Environment KOMET-PENMAN developed at GMD/IPSI (KOMET and KPML). Currently the system generates coherent text in English, German and Dutch. Extensions to other languages are under development.
This report provides a complete description of the Generalized Upper Model as released as an integral part of the KOMET-PENMAN text generation system. It includes descriptions of all the concepts defined, their interrelations, and their consequences for surface realization.
The current working definitions of the concepts are given in the knowledge representation language LOOM; there are two files: one for the upper model proper (GUM) and one for textual concepts that were previously to be found in the Upper Model but which have now been removed (Text Base).