SCOTT ME UP, BEAMIE
Ian Wishart discovers speech recognition software is light years ahead of where it used to be
It was in the mid 90s that Dragon and I first became acquainted, and I won’t say it was love at first sight. I remember walking into a Noel Leeming store and being told to come back in a week when the new, Dragon 2.0 version would be out – “much improved” on version one, the salesman assured me.
I bought it, and given the limitations of the old 486 processors (remember those?) and the software, it worked OK. Good enough to be novel, not good enough to write a novel, if you get the drift, and certainly not as fast as speaking or as fast as a good typist.
As a writer and book publisher, there were however certain tasks – like dictating someone else’s handwritten book manuscript – that made speech recognition software useful.
I vaguely recall purchasing Dragon 4.0 at some point then misplacing it, and so purchasing Dragon 7.0 a couple of years ago when, again, I needed to dictate 50 pages of transcripts of a major Maori hui dating from 1861 for an Investigate article.
It would be fair to say I was impressed by Dragon 7.0 and its ability, once I had trained it, to handle complex Maori tribal names and places like Whakarewarewa. Nothing wrong with version 7 at all and I still have it on one of the office computers, but when Mistral Software – the NZ agents for Nuance who now own Dragon – sent through the latest release, Dragon Naturally Speaking 9.0, I have to say I was blown away.
The speed at which the software now translates as you speak is incredible, and the accuracy is stunning to boot. We used it over the New Year break here at Investigate to compile the John Key/Bill English interviews in this issue – I simply played back the interviews on my MP3 recorder and live-dictated the questions and answers into the computer. The software transcribed so rapidly (Dragon boast 160 words per minute) that most of the time it was waiting for me, while I was trying to repeat a politician’s words on the wing so to speak whilst dropping the ums, ahs and pauses.
Like much of the software, and for that matter hardware, on the market today, Dragon 9.0 is capable of far much more than the average punter will use. In some programmes that’s a negative because they’re so complex to navigate and unlock. In Dragon’s case the company has always erred on the side of idiots, offering a package that allows users to either dip their toes in or plunge in. Like peeling away the layers of an onion, regular use of Dragon 9.0 turns up new tricks and new options on a semi-intuitive basis.
It can, for example, scan a downloaded voice memo from a digital recorder, Palm handheld or Pocket PC device, and automatically transcribe and type it out. Bluetooth headphone support is built in, allowing you to pace the room wirelessly while dictating. If the idea of bells and whistles excites you, why stop at dictation? Dragon is also pitching virtually keyboard-free use of your computer simply by talking to it.
There’s a memorable scene from the 1986 movie Star Trek IV: The Voyage Home, where Scottie confronts a 1980s desktop computer and tries desperately to make it respond to his voice. It was to be a further 11 years before the first Dragon programme hit shop shelves to help make Trek creator Gene Roddenberry’s futuristic vision real.
Dragon 9.0, however, is itself galaxies away from its 1997 ancestor. Nuance are boasting 99% accuracy with the latest version and, for the first time, no lengthy training process. In previous versions time had to be spent reading set scripts for yonks while the computer got used to your voice patterns. No longer. In what Nuance claim is a world first, it’s virtually out of the box and go, and the software learns intuitively the more you work with it. Streamlined processing means coughs, sneezes, ums and ahs are all screened out, so they no longer appear on screen like a swearword in an Asterix comic.
As an office tool it’s excellent. As a homework aid, it allows students to much more quickly add passages from books or encyclopedias to their work
There are specialized versions available for the legal and medical professions, and the standard version of Dragon 9.0 kicks in at NZ$259 with a “Preferred” edition aimed at small business/home business users at $449, and a Professional edition for corporate and network use available as well.