You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
59 lines
1.1 KiB
59 lines
1.1 KiB
Some measures of intelligibility and comprehension
|
|
|
|
0 10 20 30 40 50
|
|
|
|
tHfeP
|
|
S T 777
|
|
|
|
Ke[rr [ep
|
|
|
|
c EEN 2 777,
|
|
|
|
zilpp | Initial consonants
|
|
BB
|
|
|
|
SH[FF |
|
|
|
|
7T PP |
|
|
|
|
GG|TT
|
|
|
|
e
|
|
L
|
|
|
|
oo Jrr fee W
|
|
|
|
eelee v Jes V7
|
|
|
|
wWie oo |w [ne V77
|
|
|
|
N 7777
|
|
|
|
wite |-V
|
|
|
|
22(sS
|
|
|
|
0 10 20 30 40 50
|
|
Percent error
|
|
|
|
Phoneme presented
|
|
|
|
Final consonants
|
|
|
|
Figure 13-2: Distribution of errors and most frequent perceptual confusions
|
|
|
|
noted above, the pattern of errors is quite different for consonants in initial and
|
|
final positions. Such a finding is not unexpected given that different acoustic cues
|
|
are used to synthesize the same phoneme in different environments.
|
|
|
|
13.2.3 Conclusions
|
|
|
|
For the most part, the intelligibility of the speech produced by the current version
|
|
of the text-to-speech system is very high. The overall error rate of 6.9 percent is
|
|
slightly lower than the error rate of 7.6 percent obtained in the earlier Haskins
|
|
evaluation using the Modified Rhyme Test. The advantage of initial over final
|
|
consonants observed in the present study is consistent with data obtained from
|
|
natural speech by House et al. (1965), and Nye and Gaitenby (1973), although it
|
|
|
|
155
|