> Question: Is the observed pause part of the segments? or is it
> between uttered segments?
that, I think, depends greatly on the semantic one assigns to the <seg>.
As the segment is by definition rather arbitrary, it would be handy to
have a single <seg> syncing with one or more others. In our texts
<seg>s are currently exclusively used for temporal alignment.
<u><seg next="#ol2" xml:id="ol1">this</seg><pause/><seg prev="#ol1"
xml:id="ol2"> pause is also temporally aligned</seg></u>
<u><seg xml:id="ol1">this <pause/> pause is also temporally
should be equivalent from a temporal perspective. The last, in my
opinion, is even better because the pause also happens while the other
speaker speaks and not at some other point in time.
> A pause can also occur between utterances and by analogy wouldn't a
> pause internal to an utterance be one between segments?
True, one could argue that, but then of course this goes for all other
elements in model.divPart.spoken, too. As there are <u>, <pause>,
<vocal>, <kinesic>, <event>, <writing> or <shift>. Especially with a
<shift> in intonation it appears to be odd to me to split <seg>s.
Even more, when <seg>s are solely used for temporal alignment.
> <seg> elements can take "prev" "next" and "part" attributes
I know, but we tried to leave these attributes untouched, if possible,
to keep further processing as simple as possible. I thought, the spoken
features might have been left out unintentionally, otherwise it would be
probably best to stick to anchors <anchor>s for the purpose of temporal
alignment. <seg>s nevertheless, would make editing the documents easier
as ommited <seg> or </seg> tags would obviously make most xml-aware
editors complain, whereas an <anchor>s are not per-se occurring in
pairs and thus are more easily accidentally omitted.