I'm writing a JavaScript software for schools, and I'm completely new in the field of speech to text. I'd like pupils to pronounce (read) sentences in French and the software to evaluate how well it has been pronounced.

I have first been looking for an API on the cloud that analyses the sound and rates the pronounication. SpeechAce.com is doing exactly that, but for English only. Idem for Microsoft speech to text on Azure: English (an Chinese) only. I'm probably a couple of years too early.

As a backup plan, I inteded to use regular speech to text and compare the produced text to the original text shown to the student for reading. But... The same pronounciation can produce different texts (such as "Mais", "Mai", "Mes",...), especially in French => we show "Mais" to the student, and the speech to text software produces "Mes". Since "Mais" text and "Mes" text are different, my software would wrongly tell that the pronounciation of the student is wrong.

So my next idea is to produce phonetic transcription of the student voice ("m[Epsilon]") and compare it to the phonetic transcription of the text "Mais" -> "m[Epsilon]".

But is there a software (more accessible than obscure university labs) that translates speech to phonetics ?

Many thanks for your guidance.

John.

You could look at the APIs provided by dictionaries like Oxford and Cambridge as they provide pronunciation as well as being able to play individual words.

-------------------------

Thanks Mark

 

Dragon Professional Advanced Scripting/KnowBrainer Scripts
Video Examples of Coding by Voice