The usual then?
So is this down to the usual "AI" shortcomings of limited/biased training data? Or is it just phonetically really difficult to do?
So is this down to the usual "AI" shortcomings of limited/biased training data? Or is it just phonetically really difficult to do?