Comparison of Dragon Naturally Speaking and Microsoft

I just came across this recently published comparison of Dragon Naturally Speaking 8 and the dictation engine Microsoft ships with Office 2003. This is one of the most indepth comparisions of these products I've seen, and if you are about to spend a significant amount of effort training and using one of these products you'll want to read through this and understand the tradeoffs.

The main take away here seems to be DNS8 offers better accuracy than Microsoft, but also requires a LOT more CPU. The author also points out that both products do not seem to perform any better/worse on dual core machines. This sort of suggests that if you've already bought Office and are running on a low end or older PC, you might want to just use Microsoft's engine. On the other hand if your PC has ample MIPs available and you have $$ burning holes in your pockets, go and buy a copy of DNS8.

The author's findings do gel with my own personal experiences in the recent past, when my right arm was in a cast for 6 weeks. I started out with the Microsoft dictation engine and after a couple of weeks I switched over to DNS. My main motivation for switching to DNS was not recognition accuracy, but the fact that they also offered a speech interface to the over all windows desktop, which let me use the computer with minimal need to try using the mouse with my left hand. Once I got DNS trained and running, I felt it offered more accurate dictation as well, though I didn't try to objectively quantify performance as this article does.

Another thing to consider is that with Windows Vista reportedly supporting speech as a seamless platform feature vs. an afterthought, its quite likely purchasing DNS at this point could be a waste of $$. I would expect the speech engine(s) in Vista as well as how they are integrated into the desktop and applications will be significantly better than what we've seen in the past.

Read the article.

Chatter on Microsoft's Adoption of VoiceXML

There is a fair amount of commentary in the blogosphere and industry rags on Microsoft's recent embrace of VoiceXML. For example, Voxeo's Clegg Ivey views it as a confirmation of VoiceXML as THE industry standard: "When Microsoft signs on to a standard you know you're in good shape". Others are busy speculating how it finally came about, such as Robert's recent posting on Wombat Nation where he theorizes it was the Unveil aquisition last fall that brought Microsoft into the VoiceXML camp.

What makes this development particularly interesting of course is Microsoft's rigorous and logic defying attempt to squash VoiceXML in the not so distant past, with its controversial announcement of SALT. For example, recall the infamous quote X.D. Huang made: "VXML is just not technically good enough and it doesn't matter what you do. You can beat a dead horse for a long time but no matter how you beat it is still dead." It will be interesting to read Huang's interview with the InfoWorld editor next month when the Speech Server 2007 beta is released. Surely he knows by now that the only thing worse then betting on the wrong horse is betting on a dead horse! ;-)

Google's Voice Search Patent

Though perhaps eclipsed by the launch of Google's Calendar service, readers of this blog might want to check out this interesting voice search patent issued to Google this past Tuesday by the US Patent Office. Slashdot had a posting on this earlier this week with a fair amount of rather interesting/entertaining commentary.

Alcatel's Genesys Gobbles up VoiceGenie

In all the excitement this week of Microsoft finally throwing in the towel by finally getting on the VoiceXML bandwagon, you might have missed this other fairly significant development: Genesys' acquisition of Toronto-based VoiceGenie. VoiceGenie is one of the early VoiceXML pioneers. In addition to their excellent technology offerings, the VoiceGenie team has also played an active leadership role in standards and the VoiceXML forum.

Read the VoiceGenie article.

Microsoft Adopts VoiceXML!

This is not a tardy April Fools joke! The inevitable has happened - Microsoft finally embraces VoiceXML by licensing Vocalocity's VoiceXML technology for their Speech Server product. We've scratched our heads for a long time wondering why Microsoft hasn't supported VoiceXML in its Speech Server product, and its good to see it finally happen. Microsoft's joining of the VoiceXML Forum as a Promoter member is frosting on the cake.

Ironically, the very first posting in this blog (December 2004) was a tongue in cheek posting on this same topic.

Read the press release.

Voxeo: A Tough Act to Beat!

Hats off yet again to the talented folks at Voxeo! Anybody who took the time to download their recently announced Prophecy beta was certain to be impressed. As we described in a earlier posting on this topic, the Voxeo offering is the industry's first complete (ASR & TTS included) freely downloadable VoiceXML platform. Now Voxeo has bragging rights to another first in that they are the first and ONLY vendor to pass 100% of all the VoiceXML Forum's formal platform certification tests!

For those of you who are not familiar with the Forum's certification test, this might sound somewhat odd, in that you'd expect any certified VoiceXML platform (and there are many!) to pass all the conformance tests. The fact is however, that there are a fair number of optional tests in the suite that are not required for formal certification. The guys at Voxeo are sort of like the precocious geek you remember from elementary school that did all the math problems in the text book the first week of school so he could work on the interesting stuff the rest of the year when everybody else was still trying to figure out the homework assignment. :-)

Read the article.