+P Chandra Maybe now I can drop dragon speaking software to blog.Cloud processing versus local processing e.g. speech recognition
I needed to upgrade my computer to run the latest version of Dragon NaturallySpeaking. The recommended system requirements call for a Intel Pentium 2.2 GHz (dual 1.8 GHz core processor) or equivalent AMD processor, and 4 GB or more of RAM.
What are the requirements for Google’s free speech recognition? You just need an Internet connection. I’ve had no major problems dictating Gmail messages on my Nexus 10 with its speech recognition.
If you have a heftier computer system, You can change a slider in Dragon
for more accuracy, but slower performance. With Google’s speech recognition, it’s nice that the processing is done on a massive server farm that is far away. Accessibility
This is a major win for the disability communities. Nuance has somewhat of a disinterest for user groups that write open source extensions for their speech recognition program (in fairness, some of the speech extensions offer functionalities that their products deliver, except the extensions are free, and can sometimes do much more).
This is already an open source, free, speech recognition program for commands called Palaver, and it uses Google's voice APIs on the back-end. The community grew very fast, and I’m guessing it’s because the speech recognition is free.
(Dragon may still be dominant, but I see Google dictation recognition recognize some of the most obscure terms that I try to throw at it. I suppose that free speech recognition on the leading search engine, and without needing a powerful local processor = more users to submit their voice samples, and correction data).
(On a side note on accessibility, I hope Google considers working with manufacturers to put an eye tracker into devices like Android phones, and Chromebooks.
There are consumer-level eye trackers that are available now, and were demoed at CES and MWC.
An external eye tracker device can cost $99, but if manufacturers of smartphone, tablet, notebook, and laptop modify the existing built-in camera in these devices, and add an upgraded sensor, it’s supposed to only add five dollars to the manufacturing costs.
If you have a vertically propped up tablet with an external keyboard, or laptop, you could remap a keyboard button to be the “tap where I’m looking” button.
Look, touch a “single tap where I’m looking” button, look, and then touch the same button again. You don’t have to keep changing your hand and finger positions between each tap.
It would be very convenient if you could use the tracker to narrow down the context to a smaller area within your current gaze, which would limit the targets for speech to match, and consequently increase the accuracy of speech recognition.
Notably, Google has been granted an eye tracking patent that involves recording advertisement impressions through eye focus with pay-per-gaze, and another patent that demonstrates a method to unlock a device by having a sensor in a head-mounted accessory track the patterns of the pupil.
Cheap, mass-market eye-trackers would be absolutely amazing for the disability communities).