ISD-T360 ETC [List of Unclassifed Manufacturers], ISD-T360 Datasheet - Page 57

• Manufacturer Part Number: ISD-T360
• Description: VoiceDSP Digital Speech Processor with Master/Slave, Full-Duplex Speakerphone, Multiple Flash and ARAM/DRAM Support
• Manufacturer: ETC [List of Unclassifed Manufacturers]

2—SOFTWARE
• Auto-synthesized time-and-day stamp
• Support for various language and
VOCABULARY DESIGN
There are several issues, sometimes conflicting,
which must be addressed when designing a vo-
cabulary.
Vocabulary Content
If memory space is not an issue, the vocabulary
could contain all the required sentences, each
recorded separately.
If memory space is a concern, the vocabulary
must be compact; it should contain the mini-
mum set of words and phrases required to syn-
thesize all the sentences. The least memory is
used when phrases and words that are common
to more than one sentence are recorded only
once, and the IVS tool is used to synthesize sen-
tences out of them.
A good combination of sentence quality and
memory space is achieved if you take the “com-
pact” approach, and extend it to solve pronun-
ciation problems. For example, the word twenty
is pronounced differently when used in the sen-
tences You have twenty messages and You
ISD
(driven by the VoiceDSP processor’s
clock).
sentence structures:
– One versus many. (For example:
– None versus many. (For example:
– Number synthesis (English —Eighty
– Word order (English— Twenty-one
– Days of the week (Monday through
You have one message versus You
have two messages .)
You have no messages versus You
have two messages .)
versus French— Quatre-vingt ).
versus German— Einundzwanzig ).
Sunday versus Sunday through
Saturday).
have twenty-two messages . To solve this prob-
lem, words that are pronounced differently
should be recorded more than once, each in
the correct pronunciation.
Vocabulary Recording
When recording vocabulary words, there is a
compromise between space and quality. On
one hand, the words should be recorded and
saved in a compressed form, and you would like
to use the best voice compression for that pur-
pose. On the other hand, the higher the com-
pression rate, the worse the voice quality.
Another issue to consider is the difference in
voice quality between synthesized and record-
ed messages (e.g., between time-and-day
stamp and ICMs in a DTAD environment). It is
more pleasant to the human ear to hear both
messages have the same sound quality.
Vocabulary Access
Sometimes compactness and high quality are
not enough. There should be a simple and flexi-
ble interface to access the vocabulary ele-
ments. Not just the vocabulary but the code to
access the vocabulary should be compact.
When designing for a multi-lingual environment,
there are even more issues to consider. Each vo-
cabulary should be able to handle language-
specific structures and designed in a coopera-
tive way with the other vocabularies so that the
code to access each vocabulary is the same.
When you use the command to synthesize the
sentence Monday. 12:30
care in what language it is going to be played
back.
IVS VOCABULARY COMPONENTS
This section describes the basic concept of an
IVS vocabulary, its components, and the rela-
tionships between them.
P
.
M
., you should not
2-17