2021-04-07 13:16:41 UTC
Software for speech production and recognition could be an
aid for people learning a language, because they could hear
the correct pronunciation from the software, then speak it
themselves and then see whether the software understands.
However, such software itself might have issues, especially
with rarer words. For example, one program does not use the
correct sound at the end of the word (it should be an [iː])
when pronouncing the noun (plural) "minutiae", and it also
did not use the correct stress for the short word "parens"
(for "parentheses") (the second syllable should be stressed
When I spoke "phonetic" to the software, it always wrote
"fanatic". I then figured out that only one sound differed,
it was [ˈnɛ] in "phonetic", but [ˈnæ] in "fanatic". I was
pretty sure that I already pronounced this sound correctly,
but to help the software I tried to produce an e sound even
closer, viz., [ˈne], which finally made the software write
I was never able to make it write "phoneme", though. Maybe
this is a word just too special, that it does not know.
So, it seems that such software today is not always a great
help when one wants to get details correct.