The voice portion of this seems to ineffective. I noticed it seemed to be working far better then any other voice recognition program before it, so I decided to test it's ability to notice errors. It seems to always say correct even when you make obvious errors
Yes, I find it pretty lenient too. I supposed this to be intentional. The main function is served if you've practice enunciation after hearing it in a way. I'd rather be let throughw ithout explicit correction of a subtly mistake I make than to be stuck and unable to progress because I can't twist my tongue into some form demanded by an over rigorous test. I suspect there is no solution that is perfect, valid pronunications are far too diverse and and so a designer must choose to err, either on the conservative or lenient side and I aprove of the apparent choice to err on the lenient side.
I considered that, and I suppose you are right. I tried Rosetta Stone before and I am still pissed about their overly strict/ poor algorithm. They didn't remember your settings on that so I had to turn it off everytime I logged in.
My concern though is whether or not it is correcting any mistakes. I suppose since this is beta they are collecting sample pronunciation that they will use to fine tune it in the future.