Thanks to Richard for comments, which I've talked to him about already, but
I'll forward them anyway with my comments in case others were similarly
confused.
>I don't see the need to distinguish ag/AG and f/F, because you could use
>the attribute [-heavy] instead:
>
>now: IP -> (ag)0-1 (AG) AG[+head]
>why not:IP -> (AG[-heavy])0-1 (AG[+heavy]) AG[+heavy][+head]
>
>and similarly with the rule ag -> f:
>...
Yes, I've simply been using the lower case "ag" as a visually useful bit of
code for "AG[-heavy]". But in the implemented grammar this is probably a
bad idea as it makes it look like something different in kind, which it's
not meant to be, so I'll stick to the full label.
>I don't think I understand the properties of AGs: some of them look like
>properties of syllables but this doesn't make sense when I look at the
>examples. Can you explain this bit please? (Similar lack of comprehension
>of the attributes for feet.)
This is the old terminology problem: the terms "onset" and "nucleus" refer
to quite different phenomena when used at intonation level rather than
syllable level. If this means there's potential confusion when we want to
search the hierarchy for things, then it would be safer to keep terms
distinct. Two possible solutions:
(1) get rid of the redundancy in the grammar and stick with [initial] and
[final] as attributes of AGs.
(2) retain terms which are explicit about the intonational roles, and call
the attributes [INTONS] and [INTNUC] (or something).
Comments welcome about the desirability or otherwise of eliminating
redundancy, and indeed about anything else which affects the computational
usefulness of the representation.
I'll bring a revised version tomorrow, and also add in stuff I forgot about
foot position.
Jill