Discrepancies Between Database- and Pragmatically Driven NLG: Insights from QUD-Based Annotations

We present annotation findings when using an annotated corpus of driving reports as informational texts with an elaborated pragmatics for the automatic generation of corresponding texts. The generation process requires access to a database providing the technical details of the vehicles, as well as an annotated corpus for sophisticated, pragmatically motivated text planning. We focus on the annotation results since they are the basic framework for linking text planning with database queries and microplanning. We show that the annotations point to a variety of linguistic phenomena that have received little or no attention in the literature so far, and they raise corresponding questions regarding the access to information from databases for the generation process.