We had 3 audiences in mind in designing the ways in which we
tested these continuous speech dictation products:
- the possibly computer phobic executive
- the support team which would function as a help desk for
an installation just rolling out many copies of this software
to a large population
- web developers
This focus caused us to heavily emphasize the importance of human
factors in the design of these products. Be sure to see our general
test criteria outlined in the section How
We Test and the specific tests
for desktop products elsewhere on the site.
We broke down the tests into these major areas:
- installation and initial training
- dictation into the supplied word processing applet
- behavior of the text-to-speech function
- sophistication of the voice macro capability
- integration with other applications
- capabilities of the hand-held mobile transcription device
(palm recorder)
We were also interested in understanding the difference that
CPU speed might make in the performance of these products.
So we tested two configurations:
- a 450 MHz Pentium III having 128 MB of RAM (the "fast
machine")
- a 233 MHz Pentium MMX with 64 MB of RAM (the "slow
machine")
The "slow machine" was within the minimum specification for these
products. We used an Andrea NC61 microphone. We will be testing
other microphones, especially the newer desktop ones, throughout
the year.
Although all the products promise integration with popular
word processors and spreadsheets, we were more interested
in seeing how the products behaved in areas where nothing
was promised but where our audience might have reasonable
expectations. These tests included:
- AOL Instant Messenger, the online chat program
- Netscape Navigator and e-mail
- English Wizard, a data base query facility which
allows you to ask English language questions to any
data base
- Macromedia's Dreamweaver web page builder combined
with Allaire's Homesite HTML editor
Our tests of voice macro capability was designed to explore part
of the magic of speech recognition - the capacity to extend its
command-and-control capabilities over other areas of the desktop
besides dictation. To test this we tried several macros, the most
complicated of which was a parameter driven macro designed to
start Netscape Messenger e-mail program with a specific person's
e-mail address.
