Tablet PC speech recognition best practices

1 Comment

I love using speech recognition and watching whole pages of text appear right from my mouth.  The speech recognition integrated into the Tablet PC is quite good but there are some simple steps you can take to get the highest accuracy possible so I am presenting those here.  These tips apply equally as well to third party speech programs like Dragon Naturally Speaking.

Use a good noise cancelling headset for dictation.  While dual array microphones on Tablets do a decent job it is not as good as a headset.  This results in more accurate recognition results.

Train the program.  All of the speech recognition programs (including that integrated in the Tablet Edition) benefit tremendously from completing ALL training exercises provided.  Most people will immediately see greater than 90+% accuracy after all training is performed properly.

Train the program in the locale you will be doing the dictation.  Ambient background noise greatly impacts the recognition engine and it is very important to train the speech profile in the noise environment in which it will be used.  Users who want to dictate in more than one location (e. g. ofice and home) should train two different profiles in those locations with the typical background noise present. It will make a world of difference.

Error correction.  If using a program that learns through usage make sure you do all corrections through the speech interface, at least at first.  It will adjust the recognition engine and dictionary by example and end up with even greater accuracy.

1 Comment

Larry O'Brien

How important is the tone in your voice when training? When I read the training text, it’s difficult not to modulate my voice as if, well, reading a story. But when I dictate, especially technical articles, I’m much more hesitant. I certainly pause more, I’m more monotonic…

Comments are closed.