MDMA. Un modèle pour l’identification et l’annotation des marqueurs discursifs « potentiels » en contexte

Abstract

Starting from the common observation that there is no recognized closed class of Discourse Markers (DMs) and that their definition may vary from one theoretical framework to another, the aim of the MDMA project (“Model for Discourse Marker Annotation”) is to establish an empirical method for the identification and annotation of DMs in spoken French. Central to our proposal is that DMs may be described as clusters of features that, in specific patterns of combination, make it possible to distinguish between more or less prototypical uses of DMs in context. We proceeded in three steps: (i) manual identification of all so-called “potential” DMs in a balanced corpus of spoken French (5,000 words; Belgium and France); (ii) automatic extraction from the corpus of every token corresponding to the candidate DMs previously identified (1,181 tokens) ; and (iii) parameter analysis of a random sample of 200 potential DMs (syntactic, formal and semantic-pragmatic variables). The hypothesis is that the statistical analysis – based on the distributional constraints of the potential DMs at stake – should uncover a certain hierarchy between the different features under scrutiny, regarding their relevance, reliability, and generalizability (or even specificity). In the present paper, we first present the annotation procedure, then we discuss several aspects of inter-rater agreement, and finally discuss the results from the in-depth corpus-based and statistical analyses.

Keywords

• discourse markers
• annotation model
• corpus-based
• multivariate analysis
• spoken French