Then there are those words that have more than one pronunciation. For example:

wind, which can be pronounced to rhyme with skinned or to rhyme with bind. I wonder why Eloquence doesn't always seem to detect when to use which pronunciation. 

On Mon, Apr 25, 2022 at 10:58 AM, Van Lant, Robin wrote:

-
And in this case, no one has to agree with me, as there are scads of definitive references that state what I stated, I just gave one.

I won't even go so far as to say that this misspelling is uncommon, it's probably very common, but that doesn't make it either correct nor does it make the resulting word fit the phonotactic constraints of the language a given synth is using.

That's the key thing, any given synth for any given language must use the phonotactic constraints for the language and dialect it's trying to produce.  And those relate to spelling in a given language.  The 'ng' sound that is a part of any word like singing, dancing, climbing, etc., can occur nowhere but at the word final position in the English language.  You give an English language synth a surname like Nguyen (nwin - is as close as I can get to an English transliteration, and I'll bet the 'nw' that most readers understand here may still not be understood by a synth) and you will get some, shall we say, very interesting results.  For those who want to pay really careful attention, note that when you say "singing" that the central 'g' is released, but there is no equivalent release as part of an 'ng' ending.  Your tongue ends up raised at the back and you stop phonating before you release it for the 'ng' sound.

When you start mixing languages where the phonotactic contstraints differ, well . . .

Where I grew up there was a very large population that came from immigrants from many eastern European countries.  I knew someone with the surname Dziagwa.  There is absolutely no way to pronounce that correctly, or anything near to correctly, if you try to apply English phonotactic constraints to it.  The way it sounds is like the word jungle, lop off the 'le', followed by the "wa" part.  The closest I can get that might be pronounced correctly by an English synth is Jungwa.  And the 'wa' is akin to what you say if you're talking about the quick mart chain Wawa.
--