voice recognition homophones
Mark Spahn
With voice recognition, they're is always the problem
of homophones: words where there pronunciation is
the same but there spellings are different.
One solution is to proofread the text that is produced
from you vocal rendition. But to minimize or possibly
eliminate the need for a second pass through the text
produced from your speech, another solution would be
to modify your speech so that homophones never occur.
Thus, whenever you want to write "there", you say /grap/,
when you want to write "their", you say /thwin/, and
when you want to write "they're", you say /boorie/.
Here, "..." represents a word, and /.../ represents a
sound (pronunciation).
The idea is to assign each homophone-word a unique
pronunciation that is not used by any other word.
There are plenty of still-unused pronunciations to pick from,
as illustrated above.
In a one-time training session, you tell your VR software
that every time it hears /grap/, it should spell this word as
"there", and similarly /thwin/="their", /boorie/="they're".
So each VR writer develops his own set of substitute
pronunciations for homophone-words, and speaks in
a spelling-unambiguous idiolect. But what if these homophone
substitute pronunciations become standardized, so that
all VR writers use the same set of VR pronunciations?
These VR writers would constitute a subculture that
could conduct conversations in VR-speak, in which
each participant knows that /grap/ is the word "there", etc.
(Short-story writers, feel free to make of this idea
what you will.)
Question: Has anyone who uses VR already adopted
this brilliant idea? If you write by VR, how, in fact,
*do* you distinguish among there/their/they're ?
Side-question: East Asians (Chinese, Japanese, Koreans,
whose languages do not make the L/R distinction)
sometimes are unable to hear the difference between, say,
"lice" and "rice". But VR software *is* able to hear
the difference. How? Might VR be a pedagogical help
in getting East Asians to hear the L/R distinction, or
in getting gweilo to hear the different tones of Chinese?
-- Mark Spahn (West Seneca, NY)
Ray Roman
Anybody out their who uses voice recognition software? With voice recognition, they're is always the problem of homophones: words where there pronunciation is the same but there spellings are different.
One solution is to proofread the text that is produced from you vocal rendition.
-- Mark Spahn (West Seneca, NY)
Alan Siegrist
> "The said they're going to visit their parents over there."
>
> I just dictated this with Dragon Naturally Speaking without
> using any such tools.
Hi Ray,
This is interesting. Was DNS responsible for making the mistake of "the"
instead of "they" as the first word in the sentence?
This seems to confirm my experience with DNS where it seems to make mistakes
not where you would expect, namely homonyms which it actually disambiguates
well, but rather in more subtle places.
Regards,
Alan Siegrist
Carmel, CA, USA
Ray Roman
This is interesting. Was DNS responsible for making the mistake of "the"
instead of "they" as the first word in the sentence?