|
|
From text to speech: The MITalk system
|
|
|
|
|
|
and includes 55 phonetic segments, three stress markers, four types of boundary
|
|
|
indicators, and five syntactic structure indicators. It will be necessary to augment
|
|
|
the inventory of syntactic and semantic symbols in the future, but those listed in
|
|
|
the table are all that seem to be needed for a first-order approximation to an accept-
|
|
|
able prosodic pattern. An example of the use of some of these symbols to specify
|
|
|
the utterance “The old man sat in a rocker” is provided in Figure 8-1.
|
|
|
|
|
|
The old man sat in a rocker.
|
|
|
SOUND1: DH ’'AH
|
|
|
SOUND1: "OW LL DD
|
|
|
SOUND1: MM ’'AE NN
|
|
|
SOUND1l: SS 'AE TT
|
|
|
SOUND1: TH NN
|
|
|
|
|
|
SOUND1l: AX
|
|
|
SOUND1l: RR 'AA KK * - ER
|
|
|
SOUND1:
|
|
|
|
|
|
SOUND1: <EOF>
|
|
|
PHONOl: Function word: DH AH
|
|
|
PHONOl: Content word: ‘OW LL DD
|
|
|
PHONOl: Content word: MM ‘AE NN [End NOUN phrase]
|
|
|
PHONOl: Content word: SS ’'AE TT
|
|
|
PHONOl: Function word: IH NN
|
|
|
PHONOl: Function word: AX
|
|
|
PHONOl: Content word: RR ‘AA KK * - ER
|
|
|
PHONOl: Punctuation:
|
|
|
PHONQl: <EOF>
|
|
|
PHONQOZ2: Function word: DH IY
|
|
|
PHONO2: Content word: ‘OW LX DD
|
|
|
PHONO2: Content word: MM ’‘AE NN [End NOUN phrase]
|
|
|
PHONO2: Content word: SS ’AE DX
|
|
|
PHONO2: Function word: IH NN
|
|
|
PHONO2: Function word: AX
|
|
|
PHONQO2: Content word: RR ‘AA KK * - ER
|
|
|
PHONOZ2: Punctuation: .
|
|
|
PHONO2: <EOF>
|
|
|
|
|
|
Figure 8-1: Example of PHONO1 and PHONO?2 processing
|
|
|
|
|
|
8.2.1 Phonemic inventory
|
|
|
A traditional phonemic analysis of English is assumed, except for the special cases
|
|
|
listed below:
|
|
|
|
|
|
1. The diphthongs Ay, aw, ow, YU are considered to be single
|
|
|
phonemes rather than, e.g., AY = AA+YY or AA+IY or AA+IH be-
|
|
|
cause none of the two-phoneme alternatives result in particularly
|
|
|
simple rules to describe durational behavior and formant trajectories.
|
|
|
|
|
|
2. The affricates cH and JJ are considered to be single phonemes
|
|
|
rather than, e.g., CH = TT+SH for the same reasons.
|
|
|
|
|
|
3. Vowel+RR syllabic nuclei are treated internally as the special vowel
|
|
|
nuclei IxR (“beer”), ExR (“bear”), axr (“bar”), oxr (“boar”), and
|
|
|
UXR (“pure”).
|
|
|
|
|
|
82
|