I well remember my first experiences of speech-to-text software. The long training periods with new applications teaching them to recognise my voice, which because it was so different from an American voice seemed to present an almost insurmountable problem. Speaking slowly, trying to sound as BBC as possible, endless editing to correct errors, waiting for the software to catch up with my speech before uttering a new sentence. Not much fun.
Things have progressed a long way since then, and the last couple of versions of ScanSoft’s Dragon NaturallySpeaking have been delightful to use. In the old days my touch typing speed beat my speech-to-text speed hands down but now, to mix metaphors, the boot is on the other foot.
Some of that advance is due to processors moving data around a lot more quickly these days, which in turn allows software to ‘think’ faster. But there have also been a lot of improvements in the ways in which software matches speech sounds to texts or commands. With both software and hardware now at levels of ability appropriate to each other, using Dragon NaturallySpeaking really is a rewarding experience.
Nowhere is there better evidence of this than in the training requirements. Just five minutes of reading preallocated texts into the software and Dragon NaturallySpeaking 8 Preferred, the version of the software that comes in the Mobile bundle, is ready to go. You can continue training on the go and indeed after several weeks of use I am still training the software, but that first five minutes is all I needed to get Dragon NaturallySpeaking 8 Preferred recognising my speech almost as well as my mother can.
You can intersperse commands with natural talking, saying things like ‘new line’, or ‘select last sentence, underline that’ or ‘Menu, File, Save as’, or ‘Click Start, all programs, Internet Explorer’ and so on, and have the software just get on with the job. You can control the cursor and can enter text into any application that can cope with it via speech. Needless to say at this point, I am impressed with the software.
It’s not just the software you’re getting with this package though. Along with Dragon NaturallySpeaking 8 Preferred you get a Philips Digital VoiceTracer 7630 digital dictation machine. You can use this to capture audio files ready to be transcribed ‘in bulk’ at a later date. I’m not so sure about this part of the package.
The Philips Digital VoiceTracer 7630 is just a digital voice recorder and it isn’t integrated enough with NaturallySpeaking for my tastes. It records up to 17 hours and 20 minutes of speech at its lowest quality setting (10 hours and 50 minutes at medium and one hour 25 minutes at high quality), with capacity for up to 99 separate recordings in each of four folders.
You transfer clips to your PC using the supplied software and USB cable, and then need to convert each file individually from Philip’s proprietary format into WAV so that Dragon NaturallySpeaking 8 Preferred can transcribe it. It would have been nice if ScanSoft had produced some sort of automated system for handling bulk file transfer and conversion.
Still, there’s a bright side even here. Having made a few test recordings and individually converted them to WAV format they were transcribed almost as well as direct speech, even though I didn’t bother to train NaturallySpeaking to recognise recorded speech (the manual says you should set up a separate ‘user file’ for recordings).
Contact: 01483 794444