Allow specifying field values
So, identifiers say, for example, [[cid=Alice]] or [[cid:Alice]] - and then Alice optionally has attributes attached.
So now we need something similar but with a different meaning, where 'Alice' doesn't refer to anything else, but is a 'field'.
This would also allow coders to attach, for example, confidence scores to their coding.
Hmm. Seen from that perspective we could say that any code can optionally have 'field contents' attached, which can then be metadata or the information you're after (data).
[[age||39]] or [[ts||00:00:16.000]] or [[confidence||2]] or so.
If I can implement this such that the ROCK simply optionally allows these 'field values' and detaches them from the codes, that would allow implementing this without interfering with existing functionality, as you said, James.
Then the complicated thing will be how to deal these values.
Hmm, I could simply store all codes in a source that have 'field values' attached, and create dedicated columns for those that store the values.
So the example:
[[age||39]] or [[ts||00:00:16.000]] or [[confidence||2]]
would yield three additional columns in the data frame, with missing values for most utterances, but '39', '00:00:16.000', and '2' for this utterance.