> At 12:55 11/01/99 +0000, you wrote:
> >Do you want a full expression of the whole YT timing model at this stage,
> >or merely an overview of how the different parts (squish etc) fit
> >together?
>
> I wouldn't mind a copy of the code or a listing, but the most important
> thing for me at the moment is the ProXML formalism and whether it is
> capable of encoding the knowledge about timing in YorkTalk.
>
> So I want to know what kinds of contexts are used to make decisions
> about timing. This could be simply a list of contexts if you wish.
>
> Mark
>
I had a go at doing the inherent durations in a ProXML style: here's the
nucleus duration.
/* inherent nucleus duration */
for (NUC) {
if (LONG=="Y") {
if (VOC.HEIGHT=="OPEN") {
NUC_INHDUR=720;
else
NUC_INHDUR=680;
}
else
if (VOC.RND=="Y") {
NUC_INHDUR=630;
else if (VOC.HEIGHT=="OPEN")
NUC_INHDUR=640;
/* schwa is a special case; RND=="N" not included because redundant */
else if ((VOC.GRV=="Y") &&
(VOC.HEIGHT="MID"))
NUC_INHDUR=550;
else
NUC_INHDUR=600;
}
}
}
I've not formalised the onset_coda_distance in the same way, but here's
the general picture for strong syllables (which should give you an idea of
the sorts of decisions which are made).
LONG VOWELS
does Voc1=Voc2?
if yes ->
is the height close?
if yes -> OCD=140
if no ->
is the vowel [O:]?
if yes -> OCD=200
if no -> OCD=180
if no ->
is the vowel [ju:]?
if yes -> OCD=200
if no ->
is the coda voiced?
if yes ->
is Voc1 [a]?
if yes -> OCD=200
if no -> OCD=180
if no ->
is the rime [eIS]?
if yes -> OCD=250
if no -> OCD-180
SHORT VOWELS
is the coda empty?
if yes -> OCD=110
if no ->
is there a single consonant in the coda?
if yes ->
is the coda consonant a voiceless plosive?
if yes ->
is the onset empty?
if yes -> OCD=70
if no -> OCD=100
if no -> OCD=100
if no ->
is the rime voiced?
if yes ->
are both coda consonants continuant?
if yes -> OCD=150
if no -> OCD=100
if no ->
is 1st coda consonant continuant?
if yes ->
is 1st coda consonant nasal?
if yes ->
is the onset empty?
if yes -> OCD=70
if no -> OCD=90
if no ->
is 2nd coda consonant continuant?
if yes ->
voiceless plo in onset?
if yes -> OCD=150
if no -> OCD=100
if no -> OCD=100
if no -> OCD=80
So, as you can see, the YorkTalk model uses information both about the
attributes and their relative positioning (eg looking at the first CNS of
a coda cluster).
Hope this is of use.
Paul