I’ve always been bit of sceptic when it comes to speech recognition software. I had my first demonstration of its capabilities some eight years ago when it was being seriously touted as the way of the future. Who needed mouse and keyboard, when all you required was microphone and one-way conversation with your computer?
If my memory serves me correctly the product was Dragon Dictate, and when driven by grunty computer, the demo certainly impressed all who saw it. But although it was entertaining to see voice commands and dictation converted like magic into on-screen text, inevitably there was the occasional misinterpreted word and it seemed corrections were required every second or third paragraph.
Recognition accuracy has traditionally been the major stumbling block for more widespread adoption of speech recognition software – largely restricting it, in the early years, for use by people with disabilities – for whom it was marvellous and liberating tool.
Not surprisingly, over the years the number of speech recognition software programs dwindled and interest waned as other new technologies grabbed the headlines.
So when I was recently invited by Derek Austin, sales manager Australia and New Zealand productivity solutions for Nuance Communications, to see the latest version of Dragon NaturallySpeaking put through its paces, I went along to see how much progress had been made.
The first thing that impressed me about Version 9 is its ability, when combined with leading-edge headset microphone (Bluetooth wireless if you so desire), to operate in even the most extreme situations. In this case it was very busy and noisy cafe on Auckland’s North Shore.
Surprise number two was the speed and accuracy of the software. Austin gave me good demonstration of the software’s capabilities, often using combination of key-strokes and voice commands. Although there were one or two small glitches, considering the circumstances the technology is light years from where it used to be. Recognition accuracy on the previous version was already an impressive 95 percent – now it’s said to be around 99 percent.
The other great thing about the latest version is the fact that it requires minimal training for users to get up to speed – in fact, Nuance claims it’s the first desktop speech recognition software that totally eliminates the need to train the software to user’s voice. Version 8 reduced the training time from around 45 minutes to 10 minutes, and now if required, it can be skipped altogether.
The key to the accuracy of this version lies in the fact that, in addition to five different language models, it offers five acoustic models of English – UK, US, Indian, Asian and Australian/New Zealand. I had to smile with the latter option, as I’m sure if both an Australian and Kiwi dictated “fish and chips” there would probably be an entirely different outcome on screen. Still, I’m told that little ‘personal training’ should eliminate those not-so-subtle variations.
Speech recognition software already has good penetration in the education, legal and healthcare sectors where it has excellent applications in regard to teaching and dictation.
Austin says the use of the technology can also have significant impact on musculoskeletal disorders experienced by office workers and related to high keyboard usage.
Therefore you can expect to see more people talking to their PC screens in years to come – and no, I’m not writing this with NaturallySpeaking, although I have been caught verbally abusing my computer on occasion.
This latest offering from Dragon lets you add speech input and control to almost all of your Microsoft Windows applications. It integrates with Office, Outlook and Internet Explorer, as well as WordPerfect, Firefox and Thunderbird. Users can also dictate into mobile devices such as digital recorders, the Palm Tungsten and Pocket PCs for automatic transcription by the software when the device is synched with the PC. There are standard, preferred and professional versions – with the latter accessed from anywhere on network, including thin client Citrix terminals. M
Glenn Baker is regular contributor to Management.
[email protected]
