Friday, January 2, 2009

NaturallySpeaking 10 test drive

Hello.

I suppose I should actually do a blog posting about Dragon NaturallySpeaking. I've been using Dragon NaturallySpeaking, now, for about six months. The version I'm using is version 10. I got at the moment it came out. In fact I actually bought it before Amazon Canada managed to get any stock of it.

I have to say that the speech to text engine is really quite impressive. And I can easily get two hundred words per minute. I suspect a lot of the reason for this is because I speak with a fairly common northeastern American accent. A friend of mine who has a sort of Québecois French Canadian accent and he had a lot of trouble getting Dragon NaturallySpeaking to work for him. I think he had to take pronunciation courses in order to get it to work at all.

Installation went fairly smoothly. I don't really like their installer, though. It doesn't have a progress indicator on it. Well, that's not completely true. It does have a progress indicator on it but the progress indicator doesn't actually show progress. It starts at the beginning and then fills through to the halfway and then goes through to 100% and then starts over again. I didn't count how many times it did this but it was enough to be annoying. The first time the progress indicator came up I was using it to gauge approximately how long it would take to do the install. I could just hear it laughing at me when it started back at zero again. Ha ha. You thought I was done.

When you use Dragon for the first time it does sound checks and make sure that the microphone you're using is of good enough quality. I use the included Dragon NaturallySpeaking microphone and was rather surprised to find out that it failed the quality test. This turned out to be because I had a second microphone plugged into the back panel of my computer. It was using the crappy microphone that was included in the computer instead of a Dragon microphone. Doh!

My dad was curious to try it as well. Since he has a different accent from mine I was curious to know if it would work as well. At first he tried use Dragon on 64-bit Vista. It didn't work. Apparently it doesn't support 64-bit Vista. It won't even install. booo.

My dad then tried it on his desktop computer. While it installed fine, we couldn't get past the voice quality test. Unlike me he didn't have another microphone plugged in. I spent some time trying to figure out what was wrong. I could record just fine in a third-party audio application. The microphone sounded great! Unfortunately Dragon and just wouldn't work with it. I still don't understand why.

In desperation we tried it on his other laptop computer. This time it worked flawlessly. I'm still annoyed that it didn't work on his 64-bit Vista laptop or his desktop computer. I don't understand why didn't work on his desktop machine. It's most perplexing. All of this means that Dragon only successfully installed on half the computers I tried it on. That's kind of depressing. I would make sure you can get your money back just in case it doesn't work.

My dad reports that it works great on his older laptop computer. It recognized what he was saying almost perfectly. At least this means it works with a hybrid British and American accent (which is what my dad has) without problems.

Integration with random applications doesn't work as well as with the included DragonPad application (which looks like a modified version of WordPad). I'm actually okay with this. It's not a problem for me to dictate all my text and then copy and paste it where I need it. It would be nice if DragonPad would transparently save as you go, though. I've actually had Dragon crash for no apparent reason on my machine and I lost a bit of work.

That brings us to stability. The software seems kind of flaky, in my opinion. I have a fairly high standards when it comes to quality of applications. Most of the time it's not unstable enough to become a bother. The most common error I get is that every once in a while it pops up with some kind of dialogue about how it can't run the speech analyzer or something. I have no idea what the error means. If I close the application, restart it then run the speech analyzer manually it works just fine.

You do find yourself fiddling around with a microphone and other settings in order to get it to do recognize your speech better/faster. I've been having problems with Dragon inserting the word "him" every once in a while when I'm not saying anything. You may have seen this on one of my past posts. Go ahead, laugh. I don't mind. I fixed this problem by tweaking the sound driver I was using and also changing the "speed versus accuracy" slider setting in Dragon NaturallySpeaking's preferences.

The package I got was the most basic one. It didn't have things like integration with Microsoft Word. As a result I can't tell you how well it works with Microsoft Word. It works tolerably well with most applications and text entry boxes, however. Sometimes it doesn't insert a space when it should. Some other times it doesn't capitalize a letter like you would expect. Sometimes it just bugs out in a way that's so weird I can describe it. These problems don't exist to the same extent in DragonPad. I would expect that integration with word would be similar.

Dragon NaturallySpeaking does tend to use a lot of RAM. It looks like its using about 256 megs. My machine has 2 gigs of RAM. I didn't notice any problems I could trace back to lack of RAM. I'm surprised to find out that a program that uses as much ram as it does isn't 64-bit yet. On the other hand maybe I shouldn't talk since the program I work on, InteleViewer, is even more ram hungry but isn't 64-bit ready yet either.. at least we *run* on a 64-bit OS, though.

My processor is a 3.2 Ghz Core 2 Duo monster machine. Dragon runs fast on this machine as you would exepect. One thing I did notice, however, was that it wasn't threaded. If you're running on a multiprocessor machine Dragon will only use one of the processors to do it's thing. That means if you're trying to figure out if your machine is fast enough, don't count the number of processors you have. It really doesn't matter. Dragon recommends a 1Ghz CPU with SSE2 support and at least 1 Gig of ram.

My dad's old laptop uses a 1 Ghz laptop processor and, reportedly, works fine.

Dragon NaturallySpeaking definitely improves with training. Going through all the included training texts will improve accuracy. Dragon can also go through your old e-mails and documents to learn your writing style, which help quite a bit too.

The biggest problem I have with Dragon NaturallySpeaking is that I often speak too quickly. Dragon seems to be limited to about two hundred words a second. Faster than that and it tends to lose words. I tend to speak considerably faster than the average person and slowing doing to only 200 words a minute is quite an exercise in self control. Dictating at this speed give me two realizations. The first is "Wow, I speak really fast". The second is "200 words per second is insanely fast compared to typing!". Writing these blogs posts now takes a fraction of the time.

So there you have it. Apart from the odd quirk or two it actually works very well. I'm a happy camper.

By the way, this post, and the last few posts too, were dictated using Dragon.. so any mistakes you find our it's felt. :-) <= real dragon error there

2 comments:

Guillaume said...

I am dictating at around 80 words per minute now. It would seem that spending a semester lecturing has done wonder for my accent.

I think I once mentioned to you how good microphones can fail to pass the Dragon NaturallySpeaking sound quality test if the machines sound card picks up too much electric interference. Usually it is laptops that have this problem, it is peculiar that it is your dad's desktop that failed. The solution is to buy a USB headset. As soon as the AC to DC conversion occurs outside of the case of the computer there are no longer problems with interference .

Like you, I also lost work due to Dragon NaturallySpeaking dictation box. The main compatible application that I dictate to his Firefox, so installing the plug-in "it's all text" helped a lot

Andrew said...

If the problem was electrical noise, shouldn't that have showed up if I used a third party audio recording application then checked the quality of that recording? 'cause I did and I didn't notice any noise on the recording.. even with headphones.

Thanks for the firefox plugin recommendation. I'll try that.