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.
47 lines
2.7 KiB
47 lines
2.7 KiB
From text to speech: The MITalk system
|
|
|
|
the anomalous sentence test can also be compared more or less directly to data col-
|
|
lected with these same test sentences by Nye and Gaitenby (1974) and Ingeman
|
|
(1978). Such comparisons should prove useful in identifying similarities and pos-
|
|
sible differences in the speech output produced by the two text-to-speech systems.
|
|
|
|
13.3.1 Method
|
|
|
|
13.3.1.1 Subjects Forty-four additional naive undergraduate students were
|
|
recruited as paid subjects. They were drawn from the same population as the sub-
|
|
jects used in the previous study and met the same requirements. None of these
|
|
subjects had participated in the earlier study on phoneme recognition.
|
|
|
|
13.3.1.2 Stimuli Two sets of test sentences were prepared. One set consisted of
|
|
100 Harvard Psychoacoustic Sentences. Each sentence contained five key words
|
|
that were scored as a measure of word recognition. The other set consisted of 100
|
|
Haskins anomalous sentences drawn from the original list of materials developed
|
|
by Nye and Gaitenby (1974). Each of these test sentences contained four key
|
|
words. Two separate test lists were recorded on audio tape with a 3 second pause
|
|
between successive sentences. The sentences were output at a speaking rate in ex-
|
|
cess of 180 words per minute. As before, we did not correct any pronunciation
|
|
errors. Examples of both types of test sentences are given in Appendixes E and F.
|
|
|
|
13.3.1.3 Procedure = Twenty-one subjects received the Harvard sentences and
|
|
twenty-three received the Haskins sentences. Testing was carried out in small
|
|
groups of five or six subjects, each under the same listening conditions described
|
|
in the previous study.
|
|
|
|
Subjects in both groups were told that this study was concerned with word
|
|
recognition in sentences and that their task was to write down each test sentence as
|
|
they heard it in the appropriate location on their response sheets. They were told
|
|
to respond on every trial and to guess if they were not sure of a word. For the
|
|
Harvard sentences, the response forms were simply numbered sequentially with a
|
|
continuous underlined blank space for each trial. However, since the syntactic
|
|
structure of all of the Haskins sentences was identical, the response forms differed
|
|
slightly: blank spaces were provided for the four key words. Determiners were
|
|
printed in the appropriate locations in standard sentence frames.
|
|
|
|
The experiment was run in a self-paced format to provide subjects with suf-
|
|
ficient time to record their responses in the appropriate space in their booklets.
|
|
However, subjects were encouraged to work rapidly in writing down their
|
|
responses. The experimenter operated the tape recorder on playback from within
|
|
the testing room by remote control. Thus, successive sentences in the test lists
|
|
|
|
158
|