Interactional Units rely on the main following notions:
- TCU (Turn-Constructional Unit): the minimal interactionally relevant linguistic unit, which is constructed with syntactic, prosodic and multimodal resources within a given semantic, pragmatic, activity-type-specific, and sequential conversational context (Ford & Thompson 1996, Selting 2000, Clayman 2013).
- TRP (Transition Relevance Place) as defined by Sacks et al. (1974): indicates the end of a TCU and the place where another participant could take the floor; the most simple case is when turn-taking occurs, as one speaker stops and another begins. TRPs are very problematic to identify in cases of multi-unit turns, where several TCUs do not end in TRPs (Schegloff 1980, 1996; Ford & Thompson 1996), or increments and pre-beginnings, which are parts of turns but not TCUs (Schegloff 1996).
- Actions: the basic pragmatic unit, which constitutes a second criterion to identify TCUs; most part of the time action boundaries match TRPs (Clayman 2013);
To identify TCUs, comparable excerpts in German and French have been segmented both according to TRPs and to action boundaries. In a first step, this segmentation has been carried out by experts:
- TRP: identification of a moment when a TRP is reached
- Global action: identification of the completion point of a complex action, which comprises more than one local action (ex. a narrative)
- Local action: identification of a moment when an action is minimally complete (ex: the successive narrative moves within the overall narrative)
- Action type: assignment of (fairly general) action labels
- Comments: observations, insecurities, alternative codings, etc.
We identified critical cases (where TRP- and action segmentation mismatched, or where the experts disagreed), then took decisions on the best criteria to identify boundaries, particularly in cases like:
- Collaborative and delayed completion
- Choral production
- Self-repairs and specifications
- Metacommunicative items (esp. parentheticals)
- Cleft constructions
- Particles and particle combinations
We aimed to define a hybrid form of segmentation protocol that seeks to define points of convergence and divergence between interactional segmentation and syntactic annotation (macrosyntactic for French).
We noticed that the completion of actions stands more efficient than TRPs to identify boundaries in our pilot corpus, for both languages (German and French).
We are currently preparing a scientific article (position paper) about analytical discussions concerning TRPs and actions as criteria for identifying TCUs. The paper shows that these two criteria do not always converge. Results of exploratory segmentation experiences will be detailed in another contribution (upcoming paper).
Guideline (on going work)
The guidelines will present first the segmentation categories from both a theoretical and a methodological point of view. Then, they will give details on the difficulties pointed out by expert and student annotators and provide explanations and specific examples on their segmentation divergences.
The expert annotators for this annotation level were Arnulf Deppermann and Henrike Helmer for German, Heike Baldauf-Quilliatre, Véronique Traverso and Biagio Ursi for French. They established the annotation conventions and planned their implementation.
For French, two MA students, Lydia Heiden and Laurène Smykowski, tested the application of the annotation instructions and principles. They proceeded separately before sharing their results and pointing out divergent annotations.