I’ve only just seen this comparison between input methods on mobile devices, comparing speech recognition, on-screen keyboard, full-size keyboard, mini keyboard and handwriting recognition.
Guess the results…go on…
…well speech recognition came out top, with 90-98% accuracy. But doesn’t that fly in the face of conventional wisdom?
Consider that there are three main activities to entering text into a device:
- Activity 1: Think what you want to say or do
- Activity 2: Input that to the device
- Activity 3: The device recognises your inputs
Errors can occur at any point. Activity 1 is the same whatever the input method, so we can ignore it for this discussion. Activity 2 varies and Activity 3 also varies by method. Now consider typing – the traditional input method we’ve used to interact with computing devices for the last 60 years. Typing is very susceptible to errors during Activity 2, as people make typing errors, but is pretty much 100% accurate at Activity 3. Whereas speech recognition is actually much less susceptible to errors during Activity 2 (we don’t often mispronounce things), but is susceptible to errors during Activity 3.
When considering Speech Recognition people traditionally concentrate on accuracy – effectively Activity 3 – and ignore the errors made during input. In order to evaluate the overall accuracy and hence suitability for widespread use, it’s important to look at the overall level of accuracy. If you do that, maybe it isn’t so surprising that Speech Recognition comes top in this comparison.