Wednesday, September 26, 2007
Windows 2 Apples Episode 13
Good morning, good afternoon or evening as the case may be. This is Sam Caldwell with the Windows 2 Apples podcasts episode 13.
Today, I’ll revisit iListen speech to text software and reveal a compelling reason to install Vista on your Mac.
In episode 11, I offered my impressions of the Mac based iListen and compared it to DragonDictate from Nuance. Dragon sets the bar for consumer speech recognition software but unfortunately it’s not available for OS X. Chuck Rogers, the "Chief Evangelist" for MacSpeech, Inc. responded with a comment now posted in my blog. Rogers claimed that iListen "can be every bit as accurate” as Dragon but "it takes more effort to get there.”
He graciously offered to replace the defective handset shipped with my iListen software and recommended I contact MacSpeech support for tips on training iListen. In the pursuing e-mails I offered to continue training iListen. Training speech to text software requires the user read a series decidedly boring “stories” back to the program.
Nuance claims version 9 of DragonDictate requires very little training for it to be useful. This is very significant advancement given the real investment in speech recognition software is the time spent training and learning to use the product. I had invested well over a week training iListen without any noticeable improvement.
After completing all but the last training session, I reported to Rogers I saw no improvement. His e-mail reply was "believe it or not, the lack of accuracy is probably at least partially due to having read most of the training stories". My blood boiled as I read his words. I had spent a great deal of time training iListen only to discover that it was waste of time and possibly counter productive!
I expressed my anger and frustration to Rogers who replied "there is no need to get angry - it is just software." This added fuel to the fire. The major cost was not in the software but my time. To Rogers' credit he offered the services of his support team with the assurance iListen could in fact become a useful tool.
Shortly after contacting MacSpeech a new headset and USB interface arrived along with detailed instructions on how to record data to be analyzed by the Mac speech support team. The USB interface is the same device shipped with the original package. The headset appears to be higher-quality than the original.
When the replacement headset arrived, I was snowed under with work and had returned to using DragonDictate on my XP computer. Quite frankly, I was not anxious to read any more stories to my Mac but decided to give it one more chance to prove its worth. After forwarding the test data to the MacSpeech support team I was congratulated on achieving 92% accuracy and assured accuracy would be enhanced by reading the last and longest story.
Another hour spent reading to my Mac and MacSpeech support said the second test sample showed I was now achieving accuracies of 95%. I did notice some improvement and was encouraged; unfortunately, this new found hope evaporated quickly when I discovered whatever gains I made disappeared the next time I picked up the headset. Granted, DragonDictate is not 100% accurate however it is consistent in the kinds of mistakes it makes. I have not seen this kind of consistency in iListen.
I am far too spoiled by DragonDictate to continue investing more time in “training” iListen. After all, the purpose of speech to text software is to enhance productivity and I feel an additional investment of time would be wasted. I wish MacSpeech well in refining their software but after experimenting with the speech to text engine bundled with Vista I now realize I have new options.
Let me explain:
On a lark I decided to tinker with the voice recognition software shipped with Vista and to my surprise I was extremely impressed! Did I say extremely …no, let me make that profoundly impressed as in blown away. Not only is it intuitive and easy to use but it is almost as accurate as my well trained DragonDictate. I had heard this was a hidden gem in Vista and I agree.
Training Vista to type as you speak is extremely simple and for a Microsoft program the interface is uncharacteristically intuitive, beautifully designed and user-friendly. Imagine that, an intuitive Microsoft application that just plain works. I'm sure Apple users would feel very comfortable with it. One can only fantasize how the world would change if the Vista speech recognition development team were allowed to reengineer the operating system.
The sophistication and exquisite implementation of the Vista speech recognition software is only superseded by Microsoft's negligence in not taking advantage of it to promote Vista. A demonstration of the software illustrating how easy it is to take control of your PC, using a simple headset, would make one very impressive commercial. I'm sure, even if the potential customers had no intention of using speech recognition technology they would be impressed. It definitely has the cool factor so characteristic of Apple products.
If Microsoft ever decides to really market this technology, they could become the company setting the bar for speech recognition. The folks at Nuance should be sweating blood but I suspect they are relieved most Vista users don't know they no longer need DragonDictate.
I am reconsidering my options. Perhaps, I will simply purchase Parallels or Fusion and install Vista on the Macintosh. This is a particularly attractive option now that I no longer feel the need to install Dragon as well.
On a related topic:
I’ve recently discovered the New York Times technology columnist, David Pogue, felt DragonDictate was such an exceptional product that he highlighted it at the Technology Entertainment and Design conference (TED) in 2006. I highly recommend you visit the TED site at http://ted.com and watch his informative and entertaining presentation.
http://www.ted.com/index.php/talks/view/id/7
Subscribe to:
Post Comments (Atom)
2 comments:
am:
Thanks for your comments. I think your representation of my replies to you were taken a little out of context. I did send you detailed information about how ALL speech recognition programs work, which you didn't report here. I also was very clear that we are very open an honest about the fact iListen takes more INITIAL effort to achieve the same accuracy as Dragon.
In terms of accuracy, we have many customers who have actually switched from Dragon to iListen. All found that with this added effort, they achieved accuracy equal to that provided by Dragon.
My comments certainly weren't in any way designed to frustrate or anger you. Saying "it's just software" is simply that: it isn't worth getting angry over. If you have a better experience with Dragon and do not want to put forth a relatively small effort to get iListen to be as accurate, that's fine. But please do not make it sound like I brushed you off, as I certainly did not do that. We gave you the same exemplary service we give all our customers.
One last thing: let me clarify what I mean by "put forth a relatively small effort." Let's say you are getting 95-98 percent accuracy with Dragon after two days. Suppose it takes you 2-3 weeks to get to that level with iListen. From that point on, let's say you continue to use both products back and forth with approximately the same level of accuracy for the next year. Your investment to get good accuracy for Dragon was about 1% of the time you were using it, versus 4-5 percent of the time for iListen.
Put another way, this means for that first year, you achieved "acceptable" accuracy from Dragon 99% of the time and iListen 95-96% of the time. Over time, of course, this percentage becomes even more negligible.
Now, compare that initial investment in establishing a good profile to the hassle of using Windows, and wait times due to having to copy and paste or simply because running Windows on a Mac is slower. If you can honestly say its worth it, that's fine. But for most of our customers, they'd rather have the ability to not only dictate into any Mac app, but also any Windows app they might run (in Parallels or Fusion) by using a fully trained iListen profile.
In summary, spend a couple of weeks training iListen for a lifetime of excellent accuracy, or spend a couple of days training Dragon and put up with Windows, the inability to dictate into any app, and slower performance. Is it really worth it? (No need to answer - we all may have different perspectives, and that's fine.)
Chuck Rogers, Chief Evangelist
MacSpeech, Inc.
Chuck:
You seem to hang your hat on the argument that it is worth it to spend two weeks training iListen so that it can become a truly productive tool contrasted to the two to three hours at most it would take to get close to 100% accuracy using DragonDictate.
Let's do some math. Assume I'm an attorney (and I am not), charging a minimum of $200 an hour and I spent three hours setting up and learning to use DragonDictate. I believe most after those three hours, if using version 9, would get accuracies close to 99% even with no training. The attorney using DragonDictate has invested $600 of his time plus the cost of DragonDictate which I believe sells for around $100 now. He becomes considerably more productive than he was before. Total: = $700 + $70(for headset) = $770
Now let's take the attorney using iListen and give him three hours to set it up and learn to use it, and at least one hour a day for 13 more days training to get close to the accuracy experienced by my first attorney using DragonDictate. That would give us $600 to set up and learn plus one hour a day at $200 per hour for 13 more days to get iListen up to speed ($2,600). That attorney has spent the cost of your device which I believe is over $100 plus $2,600 in his time to train iListen. Let’s add now: $600 (setup) +$179 (for iListen) + $2,600 (training) = $3,379.
I believe he would be ahead of the game even if he purchased a program like parallels and a new boxed copy of Vista business and had it installed by someone from Best Buy or Circuit City
Lets say $470 for Parallels and Vista and $300 for set up which adds another $770 to our $770 yielding a total of $1,540 verses your $3,379.
I suspect you will argue that he will not have the same flexibility of entering his data directly into his Macintosh programs and that's true. That could be a make or break but I suspect for very few.
At any rate, I believe your argument does not hold water. For me I know that was certainly the case. I invested many hours, well over one week "training", iListen and many hours trying to correct it with your quirky and frequently unpredictable corrections window. As you know in my case I'm simply going to install Vista on my Macintosh and use the excellent free speech to text software built into Vista.
I will credit MacSpeech, as I have in my blog, for offering to help me and add that your company’s response time was almost immediate when I had questions or issues. However, time is money and your product was costing me a great deal of that commodity.
You also seem to ask to be excused for having a product that's not as competitive as Vista is with DragonDictate because your company is small and the Macintosh is difficult to work with. I am neither a fanboy of Windows, Apple, DragonDictate, iListen, iPhone or iPod. I use what works and gets the job done.
In today's market you simply can't afford to invest time in a product and company because you're fond of them unless that's what lights your fire. Your company did try, but my conclusion is iListen simply is not currently close to being competitive with DragonDictate nor Vista.
Post a Comment