I work on both conversational and monologic speech, and I always find it really tricky to find the "right" unit of transcription and analysis. I work on hesitation and (dis)fluency phenomena (uh/ums, silences, self-repairs, self-interruptions, vocalizations etc.)in English and French, and I compare speakers' productions in two different settings (individual oral class presentations vs dyadics interactions).
How can I find a unified way to segment both types of speech? For example, TCUs are limited to conversations since they are governed by turn-taking, so how could I apply a similar unit to the class presentations, where there is only one turn? Is it even relevant to think about a unified unit, given the differences between the two speech genres? Also, other terms such as "utterance" "interpausal unit" "intonational unit" and the like can be tricky to define, and they all have different criteria. Should I try to use syntactic, prosodic and semantic criteria to define my unit of analysis?
Thanks a lot for your help!
I don't know why you should worry about giving "a unified unit" or "unit of transcription." You are actually dealing with HESITATIONS and DIS-FLUENCY phenomena, which come in various linguistic and non-linguistic shapes and defy what you mentioned. And they are so transparent when you see or listen to your recordings or videos. I think all you should worry about is to come up, across English and French and across dialogic and monologic data, with frequencies, dominance, temporal differences, etc. across your focal phenomena and across genres if you datasets include different ones.
I hope I did not miss your point.
Clauses would be my choice, if you want to study the distribution of disfluencies across different positions: relatively easy to operationalize, and will show you interesting mid-clause uses of disfluencies. Intonation units are another option, but not super easy to operationalize (intonation contour is one way). Other options like "discourse units" or utterances, as you say, combine criteria from multiple levels so not that straight-forward, but it can be done too.
Thank you Zouheir Maalej for your answer. Finding a consistent unit for the two datasets is important to me because I am also investigating the positional features of disfluencies (on top of frequency, temporal differences etc.). So in order to compare their distribution across different positions, I need to make sure that my choice of unit is the same in the two datasets.
Ludivine Crible, thanks a lot! I have so far used intonation units (based on intonation contour) but like you said, it's not always easy. I could try looking at clauses.
Interesting issue! Perhaps have 'word' or 'speaking time' as unit of talk. They may be the best comparable across genres and modes. X pauses per 100 words, or X pauses per minute of speaking time.
Loulou Kosmala
I wonder if you have come across Robert E. Longacre’s An Anatomy of Speech Notions (1976) & Grammar of Discourse (1996). They might prove useful to your kind of research.
I measure repair fluency using the number of repairs/pauses etc per minute. Hope that helps!
- Badges
- Science topic
- Similar topics
- Academic Journals