.
/v3-uk/review/1956810/nuance-dragon-naturallyspeaking-premium-review
25 Aug 2010, Khidr Suleman , V3
Overall this is a great tool that is highly addictive. We recommend it to users who are considering making the move to dictation, and also to those who may have had bad experiences with previous versions. There are some issues with the software, but they're relatively minor.
Price: $149.99
Manufacturer: Nuance
Pros:
Quick and accurate voice recognition; compatible with multiple programs; good quality headset included.
Cons:
Resource-hungry; although some are amusing, correcting mistakes can get annoying.
Review
Voice-recognition software has come a long way in recent years, and in order to put Dragon NaturallySpeaking 11 to the ultimate test we decided to use the software to dictate this review.
The initial setup process takes 30 minutes, but it’s time well spent. Users are required to go through a set of steps including creating a user profile, which asks you to input your name, age range and whether you have an accent or not. Microphone sound quality is then checked and the final stage of the setup involves training NaturallySpeaking 11 to understand your voice.
Nuance’s claim that it can be trained in under four minutes has merit, but users are advised to spend as much time as possible training the software as this will save frustration further down the line.
However, for those who simply can’t wait to get cracking, it is possible to continue training after the initial installation; accuracy can then be improved at a later date.
Training is carried out by the user reading passages of text out loud. There are many passages to chose from, including John F. Kennedy's inaugural speech and lines from Alice in Wonderland.
A headset is included in the box and comes with an adjustable microphone. It has a snug fit, which is important considering it will be heavily used.
When dictating this review, the accuracy of the Dragon NaturallySpeaking 11 programme was quite astounding. Nuance claims a 99 per cent accuracy rate, and while we wouldn't go quite this far, it’s very impressive.
The success of the software was always going to be judged by how mistakes are corrected. Thankfully corrections are easy to make and the more the software is used, the less likely it is to make the same mistakes.
It is possible to correct a word by saying 'correct' followed by the erroneous word. The correct word can then be repeated or the user can choose from the alternatives displayed by the software. Additionally users can choose dictate the word letter-by-letter with the ‘spell that’ command; this also helps the software learn new words and is particularly useful when users want to add names of people and places.
When a word needs to be corrected but appears multiple times in the text, NaturallySpeaking neatly numbers all the words and asks the user to choose appropriate one.
Some mistakes can frustrate, but if a whole sentence is flawed it can simply be erased by saying 'scratch that'. A second attempt is usually sufficient. In our tests, the majority of mistakes actually provided light relief, injecting some humour into the most mundane subjects – albeit unintentionally, of course.
As well as speeding up the creation of documents, it is also possible to issue commands, such as ‘save the file’, ‘print the file’ and ‘open/close document’, as well as carry out common tasks in Microsoft Word such as copying and pasting.
A handy quick reference card resource is supplied with the software and lists a number of sample commands. It is a good idea to keep this close at hand while getting used to the software.
Another great feature is that the application is capable of reading back dictated work, in your own voice. This aids proof-reading since if there are any errors you have your original dictation recording to remind you what the correct word should be.
As well as dictating documents in full, the user can initiate commands including sending e-mails and dictating status updates. Browser windows can also be opened by saying a command such as "start Firefox".
Surfing the internet is particularly impressive using the software. It is possible to enter the web address and navigate to a site without having to touch the mouse or keyboard by using commands such as ‘click’ and then the link, or saying a URL followed by ‘go there’. If the software was to be used in conjunction with a laptop possessing a touchscreen, it could make using the internet a much more intuitive experience.
Nuance claims that the software is three times faster than typing. We are hard-pressed to disagree with this as the dictation we experienced was fast, accurate and saved a lot of time writing this review. We envisage it helping in countless other situations, including writing reports, speeches or essays.
As the software can be synced with Microsoft Outlook, Gmail, Mozilla Thunderbird and Lotus Notes programs, it will be an asset to administration tasks. Subject fields, recipients and the body of the text can all be inputted by voice alone.
The ability to compose emails quickly and send them without touching a keyboard will no doubt help business users save time by helping overcome a backlog of emails quickly, for example.
A new addition to the software over previous versions is the onscreen sidebar, which provides reminders of shortcuts along with tips; this automatically adjusts when you are using different programs and can be turned off when you become more proficient.
All the more impressive is the fact that the software is compatible with the majority of applications on the market today. These include Microsoft Word, Excel and PowerPoint, as well as the previously-mentioned email clients.
Other new features include the multiple suggestions offered by the correction tool and an enhanced ability to improve recognition and learn new words over time. Nuance also claims it’s 15 per cent more accurate than version 10. However, since version 10 was already 99 per cent accurate, this increase is minimal.
We do have a few gripes, though. New users will always have to set up a unique profile, so the Nuance application can’t be described as pick-up-and-dictate software. Somewhat unsurprisingly, the software is also guilty of hogging system resources as well.
We ran it on a Windows 7 laptop with a dual-core processor, and when the setup detected that we did not have 4GB of RAM it informed us that the software ‘may not perform as well as expected’.
However the speed of the dictation was still impressive, despite the test laptop only having 1GB of RAM.
There was occasional lag and, when dictating this review, the program was the most resource-heavy application, using up significantly more memory than a Firefox window with six active tabs.
This could be an obstacle for enterprise users who are looking to run the software on a relatively low-powered computer, simply because the system might not be able to handle its heavy demands.
While the upcoming Professional version will be tailored to use in a business setting, it is still hard to imagine it being used in an open-plan office environment. This is mainly because of the distraction it would cause to other staff members and the fact that it requires complete attention to use. Instead, it’s more suited to professionals who have their own office.
Nuance is clearly trying to bring speech recognition into the mainstream; it has been promoting the software in adverts on television and has also released two applications for the iPhone.
If you’re new to voice recognition, it will feel strange at first. However, it doesn’t take long to get used to and it helped us to dictate this review (well over 1,000 words) in no time at all.
Specification
Minimum 1 GHz Intel Pentium or equivalent AMD processor or 1.66 GHz Intel Atom processor.1.8 GHz Intel Dual Core or equivalent AMD processor recommended. Minimum 1 GB RAM for Windows XP and Windows Vista, and 2 GB for Windows 7 and Windows Server 2003/2008. 2 GB RAM recommended for Windows XP and Windows Vista, and 4 GB for Windows 7 and Windows Server 2003/2008 64-bit
Do you agree?
Can it really be that good?
I was one of the people who purchased the forerunner of this system, IBM ViaVoice. With every upgrade they promised 99% accuracy, as this software does, and as keen as I was to give it every chance to prove their claim, so I proved it would never do it. Admittedly the computer power has changed since those days, but I cannot forget the frustration of the odd word in 20 that was wrong and the amount of wasted time to change it using the software. I have yet to be convinced that any speech recognition software can be effective.
Posted by Trevor Clarke, 28 Sep 2010