I find the test a bit unreliable as well. Only someone well trained in giving speeches would read something twice with the exact same tone, diction and pace.
If he were to record the paragraph and play it over speakers or use that same recording twice, then we could see which system is more accurate.
They should also include a human transcription as comparison. Even when people talk to each other, often some sentences need to be repeated before they are understood.
These comparisons are disingenuous. It's much like comparing similiar cars from two different years. Plus I am sure Android will "leap frog" Siri soon anyway. With a semi open platform google gets tons of new VR apps. I have been using Vlingo for some time now and it has similar features as siri, including asking it questions and getting a response w/o going to the web and one button email dictation as in "Send email to Joe..I see you at the bar" and unlike iOS I don't have to wait for semi annual updates to get improvements.
4 comments
[ 0.18 ms ] story [ 14.5 ms ] threadI would question whether or not he had already talked to Siri before the test, as such prior experience is likely to make the whole test meaningless.
If he were to record the paragraph and play it over speakers or use that same recording twice, then we could see which system is more accurate.