The Inquirer-Home

Open Sourced IBM speech code doesn't include Viavoice

Opinion when in doubt, suspect the worst from Big Blue
Wed Sep 22 2004, 20:50
"I bought ViaVoice for Linux from IBM for about $40 in the Spring of 2002. It was the worst install I ever seen! if it wasn't for internet posts I would have never gotten it working" - Ed March
"Sadly, Viavoice for Linux has been taken off the market by IBM" - Gary Gordon
"xvoice currently depends on IBM's ViaVoice SDK (for compilation) and Runtime (for running). The SDK is no longer available. The Runtime is still available on a limited basis, but IBM has demonstrated a low commitment to speech recognition on the Linux desktop. xvoice needs a replacement that will not disappear" - the Xvoice-sphinx project.
"A group of developers and users is out there wanting to make ViaVoice for Linux a success, but without even minor support from IBM the opportunity will be missed" - Rob Spearman, October 2002

Last week we repoted, that IBM was open sourcing its Speech Recognition technology, based on a very vague story on the NY Times. First, an anonymous reader tipped me off: "That article at NYT was rather vague, and the title was quite misleading" he wrote as he proceeded to tell me that "IBM's speech recognition, text-to-speech, and other core speech technology is not involved". He continued: "Another strange aspect was that when Motorola signed on to back the effort, IBMers presented a plan of one open source project, but then announced two. It was a pleasant surprise, but does show a lack of planning".

An article published yesterday by finally shed more light on the nature and extent of the code being open sourced by IBM. In short: those of you (and me) that were expecting the opening or resurrection of the abandoned Viavoice for Linux speech recognition engine are going to be as disappointed as myself.

According to the ADT story, the code being handed to the Apache foundation is called "Reusable Dialog Components" (RDCs), which are "JSP (Java Server Pages) tags developed by IBM Research enabling dynamic development of voice applications and multimodal user interfaces". And the Eclipse Foundation is getting "speech mark-up editors designed to make it easier for developers to write standards-based speech applications". This won't be of any use for those of us expecting ubiquitous speech recognition on linux desktops.

So there go the hopes of getting speech recognition open sourced or becoming commonplace. Lesson learned: when in doubt, expect the worst from IBM. One has to wonder: is this the same company which once aired TV ads claiming to be "100% behind linux"?. And they prove that by discontinuing the Viavoice for Linux product (once available for sale), removing the Viavoice toolkit for linux from its web page, and leaving lots of people looking for non- existing alternatives?.

The IBM "response"

While the VP of Linux has remained silent lately to my mails (perhaps as VP in charge of everything-linux at IBM he doesn't have anything to say about IBM's lack of strategy with regards to speech recognition on the Linux desktop?), another IBMer named Brian Garr who works at Big Blue's "Call Center and Voice Portal Segment" told me: "We currently do not have a Linux ViaVoice SDK available" -hint: you once had, but due to IBM's strategy of being "100% behind linux" you abandoned it and pulled the download from the Viavoice SDK site!. But he continued: "We have concentrated the ViaVoice desktop brand on Windows and that product is available through our distribution partner ScanSoft. Any questions about the ViaVoice desktop product should be directed to ScanSoft". If I read that correctly, he's telling me "it's no longer our business". Strange, because at one point speech recognition was one of Big Blue's top technologies, and you know, the marketing robots at the company insist they are (I refuse to stop repeating it) "100% behind linux". So that's why you are supposed to run IBM's voice recognition product for windows, under windows emulation in linux, of course!. *cough*. He concluded "(the code open sourced) centers around the use of speech in the enterprise call center, not the desktop, which is a different product with a different set of features and functions".

Finally, the article by John K. Waters which was as pleasant as a hit in the head for me and my opensource-viavoice expectations, ends copying a full paragraph from my own, almost verbatim: "IBM is also no stranger to the speech-recognition business. The company introduced its first software-only speech recognition product, Voicetype, in 1996. It later created Viavoice, the first "continuous speech" engine. About five years ago, the company released the Viavoice Toolkit for Linux.".

Cut-and-paste journalism... we've heard of it. Remember the lesson: don't believe everything you read, not even in the INQ or the NY Times, and always expect the worst case scenario when it comes to IBM's "strategy" (or lack thereof).


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?