[DECtalk] old article about speech intelligibility

Don Text_to_Speech at GMX.com
Mon Jul 29 10:06:07 EDT 2019


On 7/28/2019 6:31 PM, Blake Roberts wrote:
> While not directly related to Dectalk, I found the article
> interesting based on the recent discussion on this list about speech
> intelligibility.

Note that it doesn't actually say anything concrete about intelligibility.
It's like grading a students math homework based on how pretty their
penmanship happens to be -- with no regard for whether their individual
answers to the MATH questions were correct, or not.  Just, how you "felt"
about the neatness of the work.

Even his subjective criteria to "evaluate intelligibility is noting reactions
to the speech" is just hand-waving.  What do you think of DoubleSpeak's
speech TODAY?  How many "oohs" and "aaws" would it get?  The speech hasn't
changed -- but the expectations as to what is possible have!

Finally, his suggestion that you effectively let users SUBJECTIVELY
"rate" it indicates that folks like me should just poll a representative
sample of intended users and see what THEY think.  While this "seems"
fair, it means that the "vote" from certain populations may be undervalued.
So, if 2.5% of the public are "blind", then their votes should only
weigh 2.5% towards the final decision -- even though they may have few
other options to process the amount of data that they'll be fed!

For example, sighted users -- who COULD rely on visual displays -- would
have more of a voice in the rating decision than folks who were more
reliant on the aural display!

Regardless, thanks for the article.  I think I'll just do what I
personally think is best and leave it to someone else to reinvent that
wheel, if they feel so inclined.


More information about the Dectalk mailing list