The first part of the sentence (in italics) is fixed, and
the second part (bold) is picked from a database. The fixed part is pre-
recorded (or canned) voice file, which never changes. However, the second
part changes from caller to caller. The account balance number is picked
from the database or calculated, converted to voice and spoken out. The
conversion for numbers, currencies, dates and characters (“ A”, “B”, “C”,
etc) can happen automatically inside the IVR. For doing this IVR system
must incorporate algorithms for the languages (e.g. English, Hindi)
that it supports. Hence you should specify the languages that you want the
IVR to support, and explicitly confirm with the vendor that it is possible.
However, few IVR’s pick text from a database
field and speak it out (e. g. weather service that accepts city STD code
from the caller, picks the text of the current weather for that city from a
database field, and speaks it out). Such systems require an additional
component called “text to speech” engine. Such engines accept text and
convert it to voice. Though this voice cannot replicate human voice, it is
quite close. These engines are generally available in English and other
world languages only. No other Indian language is supported as of now.
Example
2. You call a courier company to track your packet, and
it speaks out:
Your
packet number 123456 dated July 3 was delivered on July 5
The portions in bold are
the ones where data was picked from a database and spoken to the caller.
Example
3 . An example of TTS (text to speech) engine is if the
courier company speaks out: Your packet number 123456 dated July 3 was
delivered on July 5, and was accepted by Mr.
Murli
.
The
part in italics is text picked from a database, converted to speech and
appended to the whole voice “string”.
