Signaling coherence relations in text generation: A case study of German temporal discourse markers

This thesis addresses the question of discourse marker choice in automatic (multilingual) text generation (MLG), in particular the issue of signaling temporal coherence relations on the linguistic surface by means of discourse markers such as nachdem, als, bevor . Current text generation systems do not pay attention to the fine-grained differences in meaning (semantic and pragmatic) between similar discourse markers. Yet, choosing the appropriate marker in a given context requires detailed knowledge of the function and form of a wide range of discourse markers, and a generation architecture that integrates discourse marker choice into the overall generation process. This thesis makes contributions to these two distinct areas of research. (1) Linguistic description and representation: The thesis provides a comprehensive analysis of the semantic, pragmatic and syntactic properties of German temporal discourse markers. The results are merged into a functional classification of German temporal conjunctive relations (following the Systemic functional linguistics (SFL) approach to language). This classification is compared to existing accounts for English and Dutch. Further, the thesis addresses the question of the nature of coherence relations and proposes a paradigmatic description of coherence relations along three dimensions (ideation, interpersonal, textual), yielding composite coherence relations. (2) Discourse marker choice in text generation: The thesis proposes a discourse marker lexicon as a generic resource for storing discourse marker meaning and usage, and defines the shape of individual lexicon entries and the global organisation of the lexicon. Sample entries for German and English temporal discourse markers are given. Finally, a computational model for automatic discourse marker choice that exploits the discourse marker lexicon is presented