Discourse markers within (dis)fluent constructions in English, French and Spanish casual conversations: The challenges of contrastive fluency research

Crible, Ludivine;et.al.
(2017) International Conference on Fluency & Disfluency Across Languages and Language Varieties — Location: Louvain-la-Neuve, Belgium (15.February.2017)

Files

PascualCrible_DISF2017.pdf
  • Closed Access
  • Adobe PDF
  • 180.08 KB
Request a copy

Details

Authors
  • Crible, LudivineUCLouvain
    Author
  • et. al.
Abstract
Casual conversation is the most natural and spontaneous form of human communication, where speech is subject to the constraints of temporality and immediate communicability. These restrictions mean that discourse planning takes place as speakers go along (Briz 1998). This on-line character of spoken language (Auer 2009) is manifested in speech production and comprehension in a number of distinct markers such as repetitions, fresh-starts, unfilled pauses or discourse markers. Götz (2013) characterizes these markers as ‘fluencemes’, a term that conveys their role in contributing to the fluency or disfluency of discourse. This study focuses on one of these (dis)fluent phenomena, namely discourse markers (henceforth DMs), and examines their co-occurrence with other fluencemes across three different languages in two comparable corpora: DisFrEn (Crible under review), a dataset that contains 17,000 words of spoken conversations in English and in French; and a sample of 17,000 words from Val.Es.Co 2.0 (Cabedo and Pons 2013), a corpus consisting of spoken conversations in Spanish. We start from the hypothesis that there are recurrent patterns of sequences of disfluencies containing DMs. Our objective is to uncover and contrast these patterns in English, French and Spanish with the overall aim of determining if the DMs present in these sequences are formally and functionally equivalent across the mentioned languages. To carry out a contrastive analysis of DMs and disfluencies across languages is no easy undertaking. DMs are a complex category that groups together heterogeneous elements, for example, conjunctions (so, Fr. donc, Sp. pues), adverbs (well. Fr. bon, Sp. bueno) and verbal phrases (you know, Fr. tu sais, Sp. sabes). They are subject to a high variability in terms of their syntactic and functional status. With regard to syntax, DMs are highly flexible in terms of position to the extent that they are in many cases optional (Schiffrin 1987). DMs are, moreover, polyfunctional and context-sensitive. The same form can carry out diverse functions depending on the context or even multiple functions simultaneously. These features make DMs complicated linguistic devices that are particularly challenging for cross-linguistic studies, especially when attempting to find common ground among multilingual corpora that employ distinct theoretical frames for annotating the categories and functions of DMs (see, for example, the varying approaches pursued in the Penn Discourse Tree Bank 2.0 [Prasad et al. 2008] and in Rhetorical Structure Theory [Mann and Thompson 1988]). The lack of consensus and use of divergent annotation conventions also hinder the contrastive study of disfluencies (cf. Shriberg 1994 and Pallaud et al. 2013, among other annotation proposals). This study aims to tackle these complexities – the use of dissimilar annotation schemes for disfluencies in general, on one hand, and for DMs specifically, on the other – through the design of a method that allows a contrastive analysis of DMs and (dis)fluencies across the DisFrEn and Val.Es.Co. 2.0. datasets. Annotations in these corpora differ in three main ways: First, fluencemes are tagged in the DisFrEn corpus according to the protocol established in Crible et al. (2016) where fluencemes are annotated at word-level. In contrast, a discourse-level tagging approach is taken in the Val.Es.Co. 2.0 corpus, in which a whole fluency structure named self-repair (see Schegloff et al. 1977 and Levelt 1983) and its constituent parts were annotated in a sample of conversations (see Pascual 2016), taking as a basis the Val.Es.Co. system of discourse units (Briz et al. 2002 and Briz and Grupo Val.Es.Co. 2014). Second, for the functional annotation of DMs DisFrEn uses the interplay of four domains – ideational, rhetorical, sequential and interpersonal – and their associated functions (Crible under review), whereas in Val.Es.Co. 2.0 the annotated discourse units of the Val.Es.Co. model – specifically the subact (Hidalgo and Padilla 2006) – provide cues to identify DMs, which are associated with three possible functions: text structuring, modality and interpersonality. Third, syntactic and positional parameters of DMs are also annotated differently in both corpora. In the DisFrEn corpus, position is articulated in relation to three major units: the clause, the whole dependency structure and the turn. In Val.Es.Co. 2.0, the position of DMs is established hierarchically according to the interactional units of the discursive model: subact, act, intervention, turn, etc. The results of this study are twofold. On the one hand, we provide an account of the methodological barriers that had to be overcome in order to carry out a corpus-based analysis. We offer a number of practical solutions to the problem of mapping diverse systems of annotation in two different corpora. For instance, we found that there is a correspondence between the four functional domains for DMs annotated in DisFrEn and the three functional types of subact of Val.Es.Co. 2.0 (e.g. domains such as the interpersonal or the rhetorical one have a high correspondence with the interpersonal adjacent subact and with the modal adjacent subact, respectively). On the other hand, we identify patterns of uses of DMs within (dis)fluent structures in English, French and Spanish, highlighting the functional equivalences and non-equivalences among these three languages. The results of this study uncover recurring patterns of DM use that transcend crosslinguistic variation and that may therefore be regarded as potentially universal discourse ‘constructions’ (Fischer and Alm 2013). The study not only fills a gap in crosslinguistic fluency research, it seeks, moreover, to bring us closer to detecting universals of (dis)fluency.
Affiliations

Citations

  • APA
  • Chicago
  • FWB

Crible, L., & et al. (2017). Discourse markers within (dis)fluent constructions in English, French and Spanish casual conversations: The challenges of contrastive fluency research. International Conference on Fluency & Disfluency Across Languages and Language Varieties, Louvain-la-Neuve, Belgium. https://hdl.handle.net/2078.5/89558