The pragma-syntactic annotation, adapted from the “period grammar” approach (Groupe de Fribourg 2012), offers a segmentation of units beyond the syntactic level, for transcripts of spoken language. First, we assume that the language is articulated, with a prosodic and a syntactic level, where maximal prosodic units (usually syllables) are used to build minimal syntactic units (morphemes). The “period grammar” assumes a third level, called “pragma-syntax”, where maximal syntactic units (that we will call “clauses”) serve to build minimal pragma-syntactic units (that we will call “utterances”). Each articulation has a minimal and a maximal unit. For prosody they are respectively “phonemes” and “syllables”; for syntax the minimal unit is the “morpheme”. As such, we will now define three new units:
1) clause: maximal syntactic unit
2) utterance: minimal pragma-syntactic unit
3) period: maximal pragma-syntactic unit.
To understand what autonomy is and how it is tested, we will use a german example (for each example we give the corpus (here FOLK), the transcription name then in parenthesis the temporal boundaries in seconds):
HF: weil das (0.32) in england (0.2) mehr vorkommt oder mh (because that in england is to be found more often or mh) AJ: ja ich glaub da spielt einiges da muss man auch die biographie von dem herrn nochmal berücksichtigen aber so das war so dieses ähm es fing auch mit ner rockabilly band an (well I think a lot plays you have to consider the biography of the guy as well again but like that was like that uhm it also started with a rockabilly band) Folk_Interview (99-118).
Starting at AJ, if we test “ja”, we get an autonomous segment: AJ could stop there, it is autonomous. The next segment “ich” isn’t autonomous, but “ich glaub” is. “Ich glaub” has no syntactic relation with “ja” and prosodic cues aren’t accurate enough to allow for a revision (in fact there is a small lengthening on “ja”, hinting at a prominence), so “ja” is confirmed as a unit, a clause. Then we test “da”, until “da spielt einiges” which is autonomous. There we can revise “ich glaub” as it can be a modality of “da spielt einiges” and has a continuous intonative contour. While without those cues “ich glaub” could be autonomous or attached, those cues make us attach it to our current segment. The next segment will then revise “ich glaub da spielt einiges” to see if it is indeed an autonomous segment, a unit. A last observation in our example is “so das war so dieses ähm” where we would have four different autonomous segments: “so”, “das war so”, “dieses” and “ähm”. The prosody alone doesn’t justify attaching them together as next follows “es ging auch…” which has roughly the same prosodic cues. Instead, we can observe that syntactically, those are hesitations, reformulations on a same structure (eventually incomplete, so “cl_u”). That structure is nowhere to be found with “es ging auch…”, making it a brand new action.
Here we have a french example in a difficult situation :
spk4: voilà oui non trente-deux voilà je change vingt-trois ans (right yes no thirty-two right I’ll change it twenty-three years) ESLO_Media (1252-1254).
Prosodic cues can indicate multiples units, but are usually unreliable, unless there is a conclusive intonation. Syntactic cues are the most important in these cases, as only a new structure will be treated as pragma-syntactic. The trap is that an old structure can be amended with added elements at the front, delaying the copied part. As long as these elements aren’t autonomous, this is still considered a (micro-)syntactic phenomenon. So even if, in reformulations, we see multiple unfinished structures, without the certitude that they wouldn’t eventually repeat each other we will consider they would and count them as (micro-)syntactic. In this second example, the prosody can’t help us, not only because there is no prominence but also because the segment is partly overlapped by another speaker (as they compete for the turn, hence the repetitions). One question could be if “thirty-two right” could be a single unit, as “right” could be a modality. This could be expanded to “right yes no thirty-two right” as, again, those could simply be modalities to “thirty-two”. If the protocol has been followed, though, those should be separated: they are distinct structures and we clearly see in this context, explicitly, that “right” can be autonomous.
The annotator for this annotation’s level was François Delafontaine