If u decide to make the game I would recommend that u make it buy to play with no ads. Also make sure the pronunciation is slow and clear.

Well, I'm not in such a hurry to make this game now, but Unity's just recommended the Watson SDK (speech-to-text, among many other possibilities), and that's kind of made me excited again. Quite possible I could make something with much more value for minimal effort (in comparison to trying to implement a similar system by myself).