International audienceThe ANNODIS project has two interconnected objectives: to produce a corpus of texts annotated at discourse-level, and to develop tools for corpus annotation and exploitation. Two sets of annotations are proposed, representing two complementary perspectives on discourse organisation: a bottom-up approach starting from minimal discourse units and building complex structures via a set of discourse relations; a top-down approach envisaging the text as a whole and using pre-identified cues to detect discourse macro-structures. The construction of the corpus goes hand in hand with the development of two interfaces: the first one supports manual annotation of discourse structures, and allows different views of the texts using...