Discourse Markers and Disfluencies: Integrating Functional and Formal Annotations
Ludivine Crible · Digital Access to Libraries (Université catholique de Louvain (UCL), l'Université de Namur (UNamur) and the Université Saint-Louis (USL-B)) · 2016
While discourse markers (DMs) and (dis)fluency have been extensively studied in the past as independent phenomena, combining DM-level and disfluency-level annotations however has never been carried out before at a fine-grained level of informativeness. It is argued that the integration of formal and functional annotations, while facing a number of methodological and theoretical challenges, is not only possible and innovative (addressing the lack of consensus in the field) but also highly relevant to the investigation of form-meaning patterns. This paper reports the methodological aspects of an annotation protocol which integrates formal identification of (dis)fluency markers and a multi-layered description of discourse markers featuring, among others, semantic-pragmatic variables such as their domain and function in context. The challenges and merits of this integration are illustrated by a comparison of clustering tendencies between different functions of DMs in DisFrEn, a French-English comparable dataset. Quantitative results allow us to generate tentative interpretations of the relative fluency of some DM functions based on co-occurrence patterns, in line with a cognitive-functional approach to spoken language.