You are currently viewing our boards as a guest which gives you limited access to view most discussions, articles and access our other FREE features. By joining our free community you will have access to post topics, communicate privately with other members (PM), respond to polls, download files, upload your own photos and access many other special features. Registration is fast, simple and absolutely free so please, join our community today!
If you have any problems with the registration process or your account login, please contact contact us.
What would you do if you could cut the amount of time you spend every day typing e-mails, making forum posts, typing in reports and otherwise putting fingers to keyboard to communicate your message? For several years there have been companies such as IBM and ScanSoft whose aim has been to allow you to simply talk into a microphone and have that voice translated into text. Considering you can speak at hundreds words a minute, the potential of getting more work done in a shorter period of time is phenomenal. Writing the Great American novel while you sit back in your chair (or even lying flat on a bed) with eyes closed, hands at your side, while playing music into your headphones would suddenly become within the grasp of just about every person...literate or not.
The problem with this theory is a difficult one to conquer. The computer does not understand words and it does not understand the context of which we are speaking. And although computers do a pretty good job of figuring out mathematical formulas, displaying graphics on a screen and playing already existing sounds such as CDs and voice, the idea that it could somehow "make sense" of human words and translated into text format is a huge challenge. Perhaps, that is why I quickly became frustrated and annoyed with such programs in the past. In fact, I couldn't seem to hit the "uninstall" button fast enough.
Scansoft recently released version 8 of their Dragon Naturally Speaking program. With it came claims of "A Breakthrough in Accuracy" that provided better than 25% improvement in accuracy. But what really turned my head was their claim of a "breakthrough in flexibility" in that it was now working with Pocket PCs.
At that claim, I thought it was high time I let go of my vow to never try Voice to text apps again and give Dragon Naturally Speaking another try. Is Dragon NaturallySpeaking 8 really the breakthrough that they claim? Can it translate my deep "Isaac Hayes" type of voice into readable text? Or will the product type each one of my sentences to look something like a two-year-old would write? Read on to find out... Installing Dragon Naturally Speaking:
Installation is about like any other app. You put a CD in the CD-ROM drive and then a brings up the splash screen. You choose installation location, etc. I did have a small problem installing it, though. As seen by the screenshot below:
My problem occurred because I have Visual Studio installed on my computer. So what I had to do was go to ScanSoft's site which they have bug listed. It's a matter of deleting a certain DLL or renaming it so that it was not picked up by the installer anymore. Once that DLL is renamed the install process completed just fine. The Desktop Client
After install, you will probably start working with the desktop client. For this review I used the professional version however, all three versions of Dragon NaturallySpeaking come with a noise canceling microphone. When you first launch the program, it tests the volume level and shows you how to adjust the microphone for better recognition.
The next step is an interesting feature I have not observed in any other speech to text application. The New User Wizard will trying adapt your current speech patterns by spidering its way through your my documents folders and three year send items box in your e-mail. I if first did not think this would do anything to improve the accuracy of the speech engine, but after seeing it in action, I'm now a believer! I will discuss accuracy later, but I must say that having Dragon NaturallySpeaking pick up and spell correctly my friends names, my server names and paths on the first try -- heck it even recognized words I've made up in the past! That feature alone was just about enough to blow my mind.
Next, the New User Wizard gives you three more options of what you can do next. You can start the tutorial, see what's new in version 8 or begin dictating. "Begin dictating?", I thought. "What about the 45 minute to an hour session of reading some book so the recognition engine can translate my speech?" I figured I would start the tutorial and at that point I would have to read the book (the part of this task I was dreading).
So I launch the tutorial and much to my amazement there was no book reading! There were about eight screens of tips such as how to create a sentence with a period or exclamation point. One tape showed you how to correct misspelled words without using the keyboard and still another show you how to capitalize for words or sentences. Pretty simple stuff...
I completed the tutorial and approximately 7 minutes. There was really nowhere else to go in the program except to begin dictating. Dragon has a proprietary workpad application called DragonPad. It launched automatically when I chose the begin dictating option.
I started out typing something simple like an e-mail response to a client of mine. My first try resulted in Dragon getting nearly all of the dictation right! I had two minor edits before I hit the send button. It's spelled the names right including my difficult last name Isaac. And I am not sure, but I doubt I could've typed it as fast as I could using Dragon.
In the three weeks since I've installed Dragon NaturallySpeaking I have become even better at using it. The desktop version recognizes nearly 95%-97% of my words and allows me to type more and type faster. I answer e-mails quicker and feel as though I can construct better documents since I can get my ideas on paper faster. Desktop Client Bugs and Wishes:
One thing that can be said about Dragon is it gets better more you use it -- especially if you use DragonPad to do the majority of your dictation and edits. DragonPad picks up the nuances of the speech better, corrects easier and does things such as capitalize and punctuate sentences properly. With that said, it would be nice if Dragon was more adequately integrated into other applications. For instance, to make a post in a web forum utilizing a web browser instead of using the intermediary DragonPad would result in a higher error rate, more misspellings and uncapitalized sentences.
Another thing I have found is that it's the microphone is not just right, then more errors occur. It would be nice if there was a little more flexibility in the speech recognition so you did not have to worry so much about the positioning of your microphone. The Pocket PC client
Perhaps one of the things that excited me most about this ScanSoft Dragon Naturally Speaking before I got ahold of it, is that it has a Pocket PC client that allows you to dictate while you are driving to work or when away from a desktop computer! As a Pocket PC enthusiast, I am sure you can understand I was very excited to give this a try.
To set it up was like just about any other Pocket PC program. You install it from the desktop to the Pocket PC. Below is a screenshot of the PPC client in action. It is a recorder application, that is not unlike others on the market. There is a record, a pause a stop and a play button to it.
Another consideration for the Pocket PC version is that basically it must be imported to the desktop version to work. There is no Pocket PC translator - thus, you cannot dictate and have it translate on the Pocket PC entirely. Instead, you must rely on the desktop to do the translation. This may be a drawback for people who do not have access to a desktop PC at all times. For instance a traveling businessperson who finds themselves in hotel rooms away from desktop PCs. People who use their Pocket PC as their only means of computing will not be able to use the product.
You start the PDA version much like the desktop version by creating a new client -- the Pocket PC client:
Unlike the desktop version, the Pocket PC version requires a dictation session to train it to understand your language. There is about one half dozen selections of material you can use training - they come in the form of books that you simply read into the Pocket PC microphone. After about 15 minutes of reading, it says "You have recorded enough speech for training." So you can stop..
Loading your first training file into Dragon can result in an error if you have your training file loaded on a memory card. You will get an "unable to find valid recordings" message...but don't worry. There is a way around this.
Once you hit OK, you can browse for the training file using the drop down next to Pocket PC folder:.
As you can see now the training file on our storage card appears. Personally, I wish they would do away with that error message that is shown above. I can see that message resulting in a bunch of technical support calls.
After you choose your training file, the wizard starts adapting it to recognize your speech patterns. You simply hit the start adapting button and it takes over from there.
After that, the wizard wants to adapt your documents and sent e-mail like mentioned above for the desktop version! Did I mention this is a great feature? it really does help improve recognition believe it or not.
All that is left to do now is to dictate into your Pocket PC. Dictating to the Pocket PC is nearly the same as training the Pocket PC. Simply hit the record button and talk.
Once you are finished dictating on your Pocket PC, you must use the dragon wizard to move the recording from the Pocket PC to the desktop. You will first connect to ActiveSync then launch the dragon connection Wizard. It will find your file and transcribe it for you to DragonPad.
For me, I found the recognition to be pretty decent for brainstorming,blogging and forum posts. The Pocket PC client does not recognize as well as the desktop version, but it does a pretty good job - recognizing 90% of my words on a good day. Recognizing far less on a bad day. Also, dictating in a noisy environment adds to the error rate, so dictating in the car or my server room results in more errors.
I found the speed of the translation to be fairly decent, however for the most part, I set it to it's task and then walk away letting it work while I do other things. It is important to note that once the speech is in DragonPad any edits should take place in DragonPad. In other words you do not want to copy all of the text into MS Word and then do your editing. The reason to do all editing in DragonPad, is the DragonPad edits you make train your Pocket PC client to get better at recognizing your speech.
The Pocket PC client also requires a few tweaks or adjustments before it will work. You must adjust the gain of the microphone in your Pocket PC. The Dell Axim, which I used for most of my dictation, is not yet a ScanSoft certified Pocket PC used for dictation. The iPAQ 55xx units are supposed to get better translation than all PPCs. I did not have an opportunity to try it out on my wife's iPAQ however I do know that iPAQ sound recording via the microphone seems to be much better than on the Axim. I do not blame ScanSoft for the results of the microphone's recording quality because it is part of the Axim hardware not the software it determines the volume and quality of the recording.
I had two problems with the PPC client. One, is that when you stop a long training or dictating session, there is no way to fast forward to near the place you stopped. You basically have to listen from the beginning to re-find where you left off.
The other issue I had was the record button that starts and pauses recording didn't work on my Axim. I've been told it works on iPAQ devices. I wrote a simple Pen Macro script that worked pretty well to fix this problem for me. If you need it, it is below:
The script launches the recorder from a button press, creates a recording, pauses the recording and restarts a paused recording. Testing different Sound Bit Rates:
I thought it might be constructive to do a test of some different recording settings for my Pocket PC to see which encoding bitrates produced better/worse recordings. Scansoft has certified the Axim X3 as a "three dragon" Fully compatible with Dragon NaturallySpeaking and meets satisfactory performance and quality standards under certain operating conditions that are specifically indicated. I am using the X30 which is Dell's "bigger and better" model - though not specifically certified by Scansoft.
Anyway, to be fair, I simply read the first two paragraphs of my Tip:EASILY Encrypt Your WiFi Connection for PPC. I found the results to be surprising! See the results below: 44,100 hz, 16 Bit, Mono (HIGH SETTING) Lay we hearing or stories about the number of white open WiFi routers that can be found exploited just by a hacker driving around town with the Pocket PC, laptop or WiFi finder. This weekend was my first try at war writing, and the number of open routers I found just between my house and the local grocery store was astounding. This alerted me that in AP be needed to be put out to all my addict friends and family that you really need to lockdown at home router clinician .
Some folks have reported that they aren't encrypting their connection because it's too hard or it takes too much time. Well, sit back and relax. Take a deep breath, and we'll have you surfing over an encrypted connection in less than 10 minutes! 11,025 Hz, 16 Bit, Stereo Lately we've been hearing horror stories about the number of wide open WiFi routers that can be found and exploited just by a hacker driving around town with a Pocket PC, laptop or WiFi finder. This weekend was my first try at Lord Rye, and a number of open routers I've found just between my house and the local grocery store was astounding. This alerted me that in AP be needed to be put out to all my addict friends and family that you really need to lockdown at home router!
Some folks have reported that they aren't encrypting their connections because it's too hard or it takes too much time. Well, sit back and relax. Take a deep breath, will have you surfing over an encrypted connection in less than 10 minutes! 11,025 Hz, 16 Bit, Mono (recommended by ScanSoft for the Axim X3) Late be we've been hearing horror stories about the number of wide open WiFi routers can be found and exploited just by a hacker driving around town with a Pocket PC, laptop or white finder. The weekend was first try at or dry, and the number open routers and I don't just before House and the local grocery store was astounding this alerted me the AP be needed to be adapt to all my head friends and only that you were a the need to lockdown home router!
Some folks that reported that they aren't encrypting their connections because it's too hard or it takes too much time. Well sit back and relax. Take a deep breath, we'll have you surfing over an encrypted connection in less than 10 minutes!
Based on the results I achieved, I have mine set to the 2nd setting: 11,025 Hz, 16 Bit, Stereo. It is a good mix of low space usage and good results. My dictation above only needed about 4 minor edits. (Note each bitrate translated APB as "AP be").
To sum up the pros and cons of both Dragon Naturally Speaking Desktop and Pocket PC: Pros
Text entry as easy as talking
Most accurate speech recognition engine thus far
Ability to write voice-based macros and shortcuts
Dictate reports, e-mails, journal entries, etc. while on the road using the Pocket PC client
Gets consistently better every time you use it
No long book reading sessions for the desktop client - training sessions for the Pocket PC are short at only 15 minutes.
Pocket PC client is fast and easy to use Cons
Pocket PC client recognition not as good as desktop
May be cost prohibitive to some
Desktop version (especially Pro) has some stiff hardware requirements Conclusion
Overall, I have to say I'm thrilled that I had the opportunity to give this product another try. So far, I have just about abandoned using my keyboard on my desktop PC to do any form of writing. The desktop client takes care of spelling of difficult words so I don't have to look them up and saves time by letting me focus on what I am saying as opposed to getting it typed properly.
My trips to and from work or across town have turned into opportunities to write when I am away from my desktop computer. Although the Pocket PC version is not perfect, it does do a pretty good job of figuring out what I'm trying to say. In fact, the other day I wrote a 500 word report while in the car on the way to an appointment. The report needed minimal editing and since I was dictating, I was free to be more expressive and what I was saying.
P.S. I wrote this review using a combination of the desktop and Pocket PC Dragon NaturallySpeaking clients Score: Related Link:ScanSoft's website
Dragon v7.x was a substantial leap over the flawed v6. Our organization have used versions 4 and 5 in the past, and since v6 was so buggy, we decided to stay with v5 until v7's release. I was amazed at its accuracy... so I am looking forward to testing v8 (which will arrive shortly). The voice file that is initially created when using DNS will continue to grow the more you use it, so take care in making backups (we had machines that were hit by a virus, and hope to recover some of these voice files before reformatting). A good sound card and microphone are paramount to quality performance from the product. A clear, concise (and natural) voice is important, but shouldn't we all have learned this back in grammar school (or speech therapy)?
DNS has had an interface with a portable recorder in the past, so the inclusion of PDAs for this function was no big surprise. I would love to see a DNS app for the PDA w/o the desktop umbilical cord (it'll have to be a Lite version, since DNS requires a HIGH overhead!)... perhaps MS Voice Command? But then again, that's for another Topic thread.
__________________ James Kendrick Microsoft MVP - Tablet PCwww.jkontherun.com Lockergnome contributor- Mobile Lifestyle...using mobile devices since they weighed 30 lbs.
Dragon v7.x was a substantial leap over the flawed v6. Our organization have used versions 4 and 5 in the past, and since v6 was so buggy, we decided to stay with v5 until v7's release. I was amazed at its accuracy... so I am looking forward to testing v8 (which will arrive shortly). The voice file that is initially created when using DNS will continue to grow the more you use it, so take care in making backups (we had machines that were hit by a virus, and hope to recover some of these voice files before reformatting). A good sound card and microphone are paramount to quality performance from the product. A clear, concise (and natural) voice is important, but shouldn't we all have learned this back in grammar school (or speech therapy)?.
That is some great advice about backing up your speech files -- I am going to dedicate a rewritable CD or DVD for this purpose right now.
So do you spend most of your day using this software for your job?
Thanks for the great review. Really enjoyed hearing about the new version, makes me want to use my old version again.
The perfect mobile DragonNaturallySpeak solution for me would be to be able to use my BT headset connected to my HP2215 iPaq while driving to take down dictation. Anyone know if this is possible??
My old car is too noise for me to have the PocketPC in the dashboard cradle and use it, and I don't what to hold it close to my mouth while driving.
Many thanks James! Didn't you tell me once you use one of these speech to text products?
Brad, yes I use the speech recognition built into The Tablet OS (2005). The accuracy is pretty good but it lacks the neat productivity stuff that DNS has. I'll have to break down and purchase it one day.
__________________ James Kendrick Microsoft MVP - Tablet PCwww.jkontherun.com Lockergnome contributor- Mobile Lifestyle...using mobile devices since they weighed 30 lbs.
The perfect mobile DragonNaturallySpeak solution for me would be to be able to use my BT headset connected to my HP2215 iPaq while driving to take down dictation. Anyone know if this is possible??
It's funny you mention the Bluetooth headset. I was doing some reading about how one of the drawbacks of using a PPC device rather than a digital recorder such as the Sony or Olympus models was that you can plug a noise canceling microphone into them. And thus get better recognition. Then it dawned on me I needed to test it out with my Jabra Bluetooth headset. I'm not sure if there is any noise canceling in the headset, but I believe it picks up better than a built-in mic would.
If I get a chance, I'll give it a shot today and post my results.
Being able to use a BT headset would be a perfect combination but don't think it is possible. I looked at the BT settings on my HP2215 iPaq and it does not seem to support a headset or audio gateway profile. I wonder if other PocketPC's do?
I have 90 mins of dead time a day in the car commuting to work. I use it to listen to podcasts but being able to BE productive and "write" stuff that would be priceless.
Well there is some good news and some bad news for you. The good news is you can use a Bluetooth headset with an iPAQ 2215. It requires a registry hack to do so. The bad news is that the Bluetooth headset does not produce quality sound to the Pocket PC. I tried it earlier today and the sound was abysmal (lots of static)
Oh well, too much to expect from the HP2215 anyway. Thanks for looking in to it. Hopefully the new PPC/Phone combo's will have useable BT profiles for use with a BT headset so we can dictate while in the car.
Oh well, too much to expect from the HP2215 anyway. Thanks for looking in to it. Hopefully the new PPC/Phone combo's will have useable BT profiles for use with a BT headset so we can dictate while in the car.
I dictated a 4 paragraph blog post in my car last night on my way home. And then a 2 paragraph follow up in my car on the way to work this morning. I didn't use a headset at all. The built in mic did ok. Errors here and there, but the "correct" feature took care of most of them.
I am thinking of adding another 15 minutes of training using a different training book to my Pocket PC profile to see if it results in better transcribing. (the instructions say 15 minutes is enough) but...
One of the unwritten rules to get accurate speech conversion is to train a profile for each location that has different background noise. A lot of the effort goes into isolating what is the spoken word versus background noise. Even for the desktop version of DNS you will find that if you dictate at home and also at work, for example, if the noise level is different at the two locations then you should train a separate profile for each.
I suspect this would also apply to the PPC- speech dictated in your car would have a different noise level than say at work. This could have a big factor in the resultant accuracy.
__________________ James Kendrick Microsoft MVP - Tablet PCwww.jkontherun.com Lockergnome contributor- Mobile Lifestyle...using mobile devices since they weighed 30 lbs.
ahhh...excellent tip! I had thought about doing this, but never saw it actually suggested anywhere. My server room definately has a different "white noise" level than my car - (don't forget to turn off the music in the car either).
I have been experimenting with Vito Sound Explorer for recording on the PPC sinc DNS can convert .wav files as well. It allows for mic boost that isn't built into the Axim software. I believe it is because of the mic boost feature, I get much better transcription results from Sound explorer than the DNS app.
Another tip I stumbled upon... for better recognition results using your Pocket PC, be sure and try the different training files. I made a new profile using the Dave Barry "lost in cyberspace" training file. It blew the results of my prior profile away!
__________________ James Kendrick Microsoft MVP - Tablet PCwww.jkontherun.com Lockergnome contributor- Mobile Lifestyle...using mobile devices since they weighed 30 lbs.
I should also note that I had a conversation with the ScanSoft development team last week. And they pointed out that no matter what bit rate or stereo/mono setting you choose as long as it is 11khz or above, that it will be down sampled to 11 khz MONO. Thus, it was their opinion that recording at a higher bit rate like I did in my examples above yielded no better results than at the recommended 11 khz mono setting.
I have a question. The guide you did rocked and I bought the software. It installed to the desktop with no problem but it hasn't or won't install to the PPC...it is the preferred edition but no dice. I've uninstalled on two different machines and it still doesn't go. Any suggestions on where to find the little app?
Hi tomb and welcome to our forums!
I found that difficult to get my hands on too. It's on the CD which I don't have handy right this moment. If I recall correctly, it's in an oddly named folder like "utilities" or something like that.
Here's a suggestion. once you install it, run 1 training session only. Then let Dragon desktop adapt that file. Once adapted, I'd uninstall the pocket PC client and use a different recorder like Vito or Resco's sound recorders. Record to WAV format.
I've had far better results using WAV files than the standard dragon PPC client.