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.
Comments
Interesting. I would have expected the android phone to perform better.
I 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.
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.