An empirical resource for discovering cognitive principles of discourse organisation: the ANNODIS corpus

We describe the Annodis corpus of discourse structures for French. The corpus joins two perspectives on discourse on a variety of textual genres: a bottom-up approach and a top-down approach. The bottom-up view builds incrementally a structure from elementary discourse units, while the top-down view focuses on the selective annotation of multi-level discourse structures. The corpus is composed of texts that are diversified with respect to genre, length and type of discursive organisation. The methodology we followed involved an iterative design of annotation guidelines in order to reach satisfactory inter-annotator agreement levels. This allows us to raise a few issues relevant for the comparison of such complex objects as discourse structures. The corpus is seen as a source of empirical evidence for discourse theories, and we present the first analyses using the annotations: for instance testing hypotheses on constraints governing the structure of discourse, studying special constructs such as enumerations.