The Automated News Agency: SEMTEX — A Text Generator for German

This chapter reports on the development and current status of SEMTEX, an implemented text generator for German. SEMTEX is based on precursory work within the Japanese/German MT project SEMSYN that led to the implementation of a system capable of generating German noun groups and clauses from isolated semantic representations. Starting from this module, SEMTEX was designed to produce German newspaper stories about labor market developments. To be able to generate such texts the SEMSYN generator had to be enriched with: mechanisms for maintaining and exploiting contextual information techniques for communicating with the domain knowledge base, e.g. for the dynamic determination of grammatical tense from temporal information knowledge about abstract rhetorical schemata and the various ways for their realization within simulated political statements.