Dragon Naturally Speaking 10 Professional review
Speech recognition and speech-to-text software has improved considerably in recent years, but can the latest version of Dragon cut it in the real world?
Nuance claims Dragon Naturally Speaking 10 (DNS 10) is the world's most popular speech recognition software. This is hardly surprising, considering it has virtually no competition at all in the retail market and little in corporate sales, either. With claimed accuracy of 99 per cent, you wouldn't think there was much more room for recognition improvement, but Nuance claims a further 20 percent, so you do the maths.
Initial training takes about five minutes and you read reasonably entertaining extracts from a choice of books to give the application a feel for your speaking style and accent. DNS 10 includes pre-defined style settings, but all these are for US regions, rather than Yorkshire or Somerset.
Nuance has updated the icons in its interface, but it's a shame it couldn't have gone the extra mile and given the whole interface a facelift. It's beige and white bars and dialogues haven't changed since the days of Windows 95 and the Accuracy Centre, with its ugly Times text and messy layout, would shame many a Freeware author.
Cosmetics are one thing and the functioning of the program is another. It's hard to test overall accuracy, as the software improves with use, by homing in on your particular vocabulary. You can give it a head-start with this, by letting it look through your document and email folders and compile a personal word list.
Initial recognition tests were good, but some way off the one word in a hundred the accuracy headline suggests. A lot depends on the quality of the sound the software hears. A noise-cancelling, cabled headset is provided with the program, but dedicated wireless headsets and some Bluetooth devices can also be used. A cabled installation usually gives better accuracy, but may not be as convenient, if you need to move away from your desk while dictating.
Nuance claims a speed increase of 50 percent in recognising what's been said, and in our tests it's noticeably quicker than version 9. If you speak phrase by phrase, as many people do, it's useful to be able to see the recognised text come up as you're speaking.
In This Article
Preparing for AI-enabled cyber attacks
MIT technology review insightsDownload now
Cloud storage performance analysis
Storage performance and value of the IONOS cloud Compute EngineDownload now
The Forrester Wave: Top security analytics platforms
The 11 providers that matter most and how they stack upDownload now
Harness data to reinvent your organisation
Build a data strategy for the next wave of cloud innovationDownload now