Write what is spoken
I'm finding that the computer generated voice is little on the weak side. Half the time I haven't a clue what is being said because it all comes through garbled. I've been studying French for awhile, and this is probably true of any language, being able to comprehend French being spoken is huge. Also the voice recognition thing doesn't really seem to do that great of a job. I notice this when I totally messed up the phase I was supposed to speak and it said I was correct. So I tried saying something totally different than what I was supposed to say, it took that as correct as well.
The system is cool and fun and all, it's fun and helps with some supplemental training along with other sources, but until they get the speech parts worked out, it's a fail.
I honestly think that Duolingo should simply rely on native speakers to record audio samples. A voting system where people would choose the best audio could avoid having useless translations. The staff could ultimately double check the best sentences and tadaaa, we have near perfect audio in Duolingo !
I read somewhere on this forum that very early in development, Duolingo tested the robot voice against actual speakers using A/B testing. Robot voice won out. So I'd say we aren't likely to see a return of native speakers.
Oh that's funny, at about the same time too. I just checked your thread, you got my up vote ^^.
I would imagine that Duolingo users would happily volunteer to record phrases in their native language. I know I would!
I agree. I have missed a few Spanish translations because the computer voice seems to become softer on certain words/phrases. And the pronunciation difference of nosotros/nosotras is so slight and almost slurred by the computer voice, that I often type the wrong one and am marked incorrect. I know the difference; I'm just not hearing it.