Some preliminary comments follow, based on your document, which is
admirably lucid, if challenging to those of us (i.e. me) who are new to
this kind of representation. (For instance, am I right in thinking that
"inheritance" works from the bottom up, rather than trickling down from
higher nodes? Like voicing can be "inherited" by the rhyme from the coda?)
* Strict binarity in onsets and codas: implications.
I am entirely sympathetic with the theoretical arguments (though I have
forgotten how current phonological theories deal with -CCC codas like
"sculpt"). In phonetic interpretation, how much does it buy you? It
allows you to relax certain phonotactic constraints (e.g. in super-heavy
rhymes), and to predict the voicing values of suffixes etc -- but this is
all information which is already encoded in the dictionary's phonetic
representation, so in that sense you don't need it. Having inflexional
suffixes assigned to an appendix, or at least marked with some kind of
[+suffix] attribute, makes long-term good sense (could be used to capture
timing differences in Scottish English), though I'm not convinced Southern
British makes pronunciation distinctions. But I'm a bit worried if we
don't have some way of incorporating these things into structure; it's a
bit like my phrase-initial bits which aren't "really" part of AGs/feet.
They are so common I don't feel I can afford to leave them out of the set
of structures we look at. Polymorphemic words (where the no. of morphemes
> no. of syllables) are pretty common too! Can we afford to exclude them?
(This is not my patch of course, so forgive me if I am simply suffering
from comprehension-deficit.) NB: the syntactic parse (Alex's phase1.syn)
gives explicit info about plurality, tense etc -- can this be exploited in
our co-referencing?
Your complementary distribution of /s/ and asp. in onsets is an ingenious
trick. Do you end up having two kinds of /s/, one for clusters and one for
singleton /s/ (which must be further specified to distinguish it from /f/
and /T/)? Do you need and/or want quite different structures for /sl/ and
/fl/?
*Ambisyllabicity.
Should be something we can handle.
I don't know whether Alex feels he now has sufficient knowledge to attempt
a parse into syllables and component parts????? We're moving in that
direction, which is good.
* Larger units.
I'll look forward to further comments on identifying (or even allowing) s s
feet.
You'll have seen from Alex's prosodic parse (phase1.prs) that we've already
done as you suggest about the phrase-initial bits, and incorporated them
into a special type of AG and foot, which can be discounted when we only
want to look at those with a proper head. AG attributes are things I still
have to spell out; <type pre-initial> (!) might be better than <type
initial>, which I might want to assign to the first *real* one in an IP...
* My Group 1 data for Phase 1 DB
No problem to replace polymorphemic coda structures if that's what you want
(see above!). For consistency, should one also treat non-cluster suffixes
as "special", e.g. "mislays", "mislaid"? I'd prefer to supplement rather
than replace, but don't want to extend the DB unnecessarily.
>
>John is away all week, so I haven't talked this over with thim. He might
>want to do one or two things differently, but basically I think this is
>something we'd both be able to assent to.
>
>If you need help interpreting what I've written (parts are a bit
>technoid), let me know.
>
>Have a good weekend!
>
Have a good one yourself! And thanks again for the response.
Jill