Top critical review
37 people found this helpful
An overrated program with a lot of potential but a lot of rough edges too
on April 3, 2011
1) capable of decent accuracy with modest training, and very good accuracy with more extended training
2) can keep up with relatively fast speaking meaning overall excellent speed of text generation
3) capacity to train multiple users and work within multiple dialects
1) unreliable, crash prone, and buggy - this program needs major work
2) poorly sorted out Windows 7 correction dialog operations
3) program becomes unstable when microphone is turned off for protracted periods of time, requiring reinitialization.
4) program support is a joke
It's almost impossible to be on the Internet or working with the computer and not be aware of Dragon's NaturallySpeaking 11.0. You would think from the relentless advertisement by Nuance that the program never makes a mistake, and allows seamless control of your computer while allowing you to dictate virtually everything. Unfortunately, the program's actual functioning falls well short of the hype and promise.
Although quite good accuracy is possible with modest levels of training, and very good accuracy is possible in technical fields with training of technical vocabulary, the program's buggy operation, recurrent vulnerability to crashing, humongous speech files and subsequent painfully slow loading, and very slow correction interface (much lower for some reason on Windows 7 than on Windows XP) spoil an otherwise enjoyable experience badly.
Although I am almost totally dependent on the program for professional writing (cranking out roughly 20 single spaced pages a day in a scientific area), I am continually amazed with how many rough edges the program has that Nuance seems to be oblivious to, simply because they recur from edition to edition. As but a small example of lack of attention to relevant details, the program, within Microsoft Word, does not use the specialized characters for quotation marks, but uses a generic and plainer character set, instead of the one generated within Microsoft Word. This means that if you want professional looking final product, you have to run a macro to replace their generic quotation marks with the proper special character. Although this is a small issue, it's just an example of so many ways in which the program falls down, and displays its lack of polish. It is amazing that despite five editions of the product which I have used that this particular issue continues, despite I'm sure much feedback from multiple users.
Then there is the issue of loading and speed of operation. With a solid-state drive (which gets to the Windows 7 desktop in 20 seconds), it takes 25 seconds to load speech files on NaturallySpeaking 11.0. In other words, an extremely fast hard drive loads a bloated operating system faster than it loads NaturallySpeaking 11.0. I hesitate to think how long it would take somebody to load speech files on the average 250 GB 5400 RPM hard drive. You could go out, mow your lawn, calculate your taxes, then do your email, and come back just in time to start dictating.
Well, so what? Once the program is loaded, things should be okay, right? Well this is where again the program falls down. First of all, actually correcting any text reveals a major bug in their Windows 7 code around keyboard input within the corrections dialog box. When you bring up the correction dialog box, you can't type characters without generating a long delay before the program somehow incorporates your keyboard input (roughly 8-9 seconds) - this is unacceptable waiting time for entering a couple of characters. Dictating or spelling out words within the correction dialog box produces somewhat snappier results (and if you disable speech playback, things are much quicker), but is still amazingly slow (running this program on an extremely fast processor, with 6 GB of RAM, and the fastest 250 GB solid-state drive). It's scary to imagine how slow this program would be on the average system. In fairness, some of it appears to be a Windows 7 incompatibility, as I use the same speech files and the same 11.0 version on a Windows XP system, and had much snappier response from within the corrections dialog box. Nuance advertises that their code is Windows 7 compatible - so something got lost in the translation. But again, this suggests that Nuance is more concerned with their advertising than with the program's actual performance.
The biggest problem though is what happens when you put the program into 'Suspend' by turning off the microphone - and by the way, you are forced to do this because if you simply tell the program to "go to sleep," any sound into the microphone will cause the page to scroll back to where the cursor currently is sitting. In other words, any kind of sound acts as a 'reinitialization command' to the cursor - the logic behind that completely escapes me! So essentially you are forced to suspend microphone input if you want to scroll through your text instead of the program intrusively interfering with this. This is where things get really interesting (after you turn the microphone off). After about 5 min., the program will malfunction (again only in Windows 7 and not in Windows XP for some strange reason) and give you an error code which the program cannot recover from, and thus all this nonsense requires that you basically start from scratch all over again. This means not simply reloading the program, because the program actually doesn't pull itself out of memory reliably and oftentimes leaves a huge executable sitting in the background (taking up over 300 MB of RAM) - - if you don't kill the program this way, when you try to reinitialize it, you will immediately get the same error code. This means you have to bring up Task manager, find the program, and manually terminate it, and then start from scratch and reload your speech files. This will again take a long period of time - close to 30 seconds with a solid-state drive, a minute and a half with a slower, conventional hard drive. All of this nonsense of course significantly deteriorates the productivity advantage associated with dictating versus typing. Indeed, for short pieces of text, all these hassles mean that it is almost always faster to simply type out a couple of paragraphs then it is to struggle with the program. For longer text generation, and assuming that you dictate continuously and don't risk putting the microphone into the off position for any extended period, the program is obviously much more productive and faster than typing.
This process of recurrent shutdown and reinitializing (due to fatal program errors) can be very frustrating, and oftentimes has to be repeated multiple times, because if you leave the program on (by not turning off the microphone), and there is any background noise, you'll get endless junk words into your dictation space. This brings us to another striking failure of the program - its inability to discriminate speech from background sounds and non-speech noises. The program seems to believe that every conceivable sound must translate into a word, and therefore, any kind of background noise makes it impossible to use the program particularly if you are not using a noise canceling microphone, and instead are dictating on a laptop that has a standard integrated microphone. You would think that the program, which runs on an enormous number of laptops, would be optimized for the standard integrated microphones on laptops. Unfortunately, it is not. You have to use Dragon's headphone microphone gear or some equivalent. Curiously, a much earlier version of the program NaturallySpeaking 8.0 seem to work much better with laptop microphone arrays. Why this might be is beyond me.
Because the program code is buggy and unreliable and subsequently the program locks up, you will find that occasionally your speech files are completely trashed and cannot be repaired through Nuance's repair algorithms (there is a speech file 'Restore' function from within the program). This means you had better make a backup copy of your speech files (make that several backup copies), or you will find yourself having to retrain your user identity all over again, from scratch - if you have spent years training various version of the program, this loss of speech files without a backup could make you suicidal or homicidal!! Additionally, you will find that their optimization program (which gives you the chance to optimize speech files that have had numerous corrections made to them) is not reliable and also occasionally locks up.
In case you're suspicious that these difficulties are unique to my machine, or are a function of viruses, spyware, or other hardware problems, rest assured that none of these issues are remotely responsible for the Programs buggy and unreliable operation. I run two spyware and antivirus programs concurrently, my system is about as fast as a window 7 system can be (Windows experience index 6.7 overall), and I always keep the machine in good working order.
All of these operational headaches spoil what otherwise might be a useful program. I believe that all of this emerges from Nuance having a very grandiose notion about their software, and not really listening very much to end-users. It's a shame that there is no real competition in the PC marketplace when it comes to dictation software. If there was, I suspect Nuance would have to get their act together because they would quickly lose market share. Given all these issues, 2/5 stars is actually generous - 1/5 stars would be justified.
I wish I could make do without the program, or transition to another speech recognition platform, but unfortunately I can't.
Well, Nuance seems to have been listening, but of course they never admit that their program has any real flaws, so they quietly released 11.5 - and it is somewhat better than the very buggy 11.0. It is less likely to lock up, although Microsoft Security Essentials doesn't seem to like this program, and sometimes you will have to stop MSE from running to dictate (perhaps MSE regards this as a virus!) Corrections are reasonably snappy, if you disable playback, or curiously, use an external mic. Not clear why the choice of microphone effects speed of operation of the corrections dialog box but it does. Nuance's support is still terrible. Improved product, but still 2/5 stars, but edging towards 3/5. No sign that Nuance is truly as committed to product debugging though as to advertizing.
Maybe next year?