from-text-to-speech-the-mitalk-system/pages-txt/205.txt

Context-dependent rules for PHONET

. +YU +alveolar, +plosive
Diptar(f2] T 200/ [ _] [ $ ]
F2 time constant longer in +F2back, +round transition:

Tef[£2] 1 50/ [ +fZIS)ack] [ +rg1ind]

Y-glide not as dramatic before alveolar consonant:
- +yglide +alveolar
Diptar{f2] { 150 / [ yeu ] [ S ]

Formant centralized in a short nonretro vowel: Shift Target[fl..£3] and
Diptar[f1..f3] towards 490,1450,2500 Hz by an amount which varies as a dying ex-
ponential with time constant 60 msec in the duration of the segment.

Schwa takes on formant targets of adjacent segments:

Target[f3] « average(Oldval[f3], Target[f3], Nextar{f3])/ [ +schwa]

Back cavity resonances assume target quickly:
Tof[£2] | 20/ [ +f2_‘f°k]
C.7.1 Diphthong rules
/[ +diphthong]

Set Tediph and Tdmid from Table C-5.
Tcenter[f1..f3] « Tdmid * Segdur{fl]
Tcdips[f1..f3] « Tcdiph * .5 * (1 + Segdur{f1] / 45)

Earlier diphthongization after HH in tense vowels:

Tcenter[f1..£3] « Tcenter[fl..f2] * .67 / [ +§H] [ +E_X ]

Bvalf(f1..f3] « average(Target[f1..£3], Diptar[f1..f3])

Bvalf[f1] T 70/ [ ‘ffi‘]

Bvalf[f2] T 200/ { [Jf]/[t(f] }
Bvalf[f3] 200/ { [=]/[=] }

193