On the signalling of multi-level discourse structures
Résumé
We present a top-down approach to discourse annotation which considers the text as a whole as a functional and semantic unit and focuses on the identification of high-level textual patterns. Two discourse objects are chosen to bootstrap this global approach to discourse: enumerative structures and topical chains, envisaged as "neutral" structures covering a wide range of discourse organisation phenomena and realised via diverse textual patterns. We present here the NLP-assisted annotation method that has been carried out, involving automatic tagging and pre-marking and the use of a purpose-built annotation interface.