|
|
Preface
|
|
|
|
|
|
and developed new testing methodologies, and provided a systematic assessment
|
|
|
of MITalk’s output speech quality.
|
|
|
|
|
|
Throughout all of the research, many important individual projects were com-
|
|
|
pleted which focused on issues in speech analysis and processing, and in linguis-
|
|
|
tics. The many participants in these endeavors focused individually on a variety of
|
|
|
important research issues, but they also shared in the motivation provided by the
|
|
|
goals of the overall system, as well as in the daily interaction with others involved
|
|
|
in complementary aspects of the system. This tension between individual research
|
|
|
and overall system building evolved with MITalk, and provided each contributor
|
|
|
with a strong sense of satisfaction derived not only from individual efforts, but also
|
|
|
from the system’s overall achievement.
|
|
|
|
|
|
In the summer of 1979, it was felt that the MITalk system was at a suf-
|
|
|
ficiently complete state that a specialized, intensive course devoted to its exposi-
|
|
|
tion was appropriate. Accordingly, from June 25th through June 29th, a special
|
|
|
short course was offered. Lectures covered all modules of the MITalk system, and
|
|
|
laboratory exercises combined with demonstrations provided further contact with
|
|
|
the system. The individuals involved with the course included J. Allen,
|
|
|
D. H. Klatt, M. S. Hunnicutt, R. Carlson, B. Granstrom, and D. Pisoni. In ad-
|
|
|
dition, a set of notes for this course was developed. M. S. Hunnicutt wrote the
|
|
|
sections of the notes covering text preprocessing, morphological analysis, phrase-
|
|
|
level parsing, morphophonemics and stress adjustment, letter-to-sound and lexical
|
|
|
stress, and fundamental frequency contour generation. D. H. Klatt wrote the sec-
|
|
|
tions on speech synthesis technology and the Klatt formant synthesizer.
|
|
|
D. H. Klatt, R. Carlson, and B. Granstrom wrote the sections on the phonological
|
|
|
component, the prosodic component, and the phonetic component. D. Pisoni wrote
|
|
|
the section on measurement of intelligibility and comprehension directly
|
|
|
reproduced as Chapter 13 of the present volume. J. Allen provided the introduc-
|
|
|
tion, a section on implementation, and the summary. These notes have constituted
|
|
|
the most comprehensive overview of MITalk until the publication of this book.
|
|
|
|
|
|
Since 1979, the MITalk system has been available for license, and has been
|
|
|
acquired by many industrial firms and universities. Bell Northern Research ac-
|
|
|
quired the system for research purposes and recoded it in VAX-VMS PASCAL.
|
|
|
They have kindly supplied a copy of their version to us. In turn, this version was
|
|
|
converted to run under Berkeley 4.2 BSD UNIX, using the syntax of Berkeley
|
|
|
PASCAL, although some routines in the new version are written in C. This latest
|
|
|
version was accomplished by R. Armstrong, and it has many new features. The
|
|
|
most notable feature is the overall control structure which easily permits assem-
|