CNET también está disponible en español.

Ir a español

Don't show this again

Culture

IBM overhauls voice-recognition strategy

Big Blue decides to shift its focus from individual products to delivering core technologies, a nod to the increasing popularity of small devices that demand voice-recognition capabilities.

    IBM tomorrow will completely revamp its voice-recognition technology strategy, focusing more on delivering core technologies than on individual products.

    The shift recognizes that computing is increasingly moving to smaller devices, such as handheld computers, cell phones and wearable PCs, that demand voice-recognition capabilities. IBM is also betting there will be big demand for voice-recognition server software essential for other devices, such as set-top boxes and in-vehicle Web appliances.

    Although IBM has a head start on competitors and a good vision for enabling voice-recognition technology in new places, analysts warn the company still has a long way to go.

    Big Blue's latest effort "points to future successes in this space," Technology Business Research analyst Bob Sutherland said. "I don't know how many clients will adopt it at this point, but it will be interesting to see."

    Until 1997, IBM focused more on the shrink-wrap ViaVoice product as it tried to create more interest in speech recognition.

    "The last few years we have been driving it toward the enterprise as a technology," said W.S. "Ozzie" Osborne, general manager of IBM Voice Systems. "What we're trying to do is build an end-to-end distributed platform and tools for people to create speech applications."

    In the same way that Microsoft worked to get software developers to standardize its tools for writing new applications, IBM is positioning its "tools and technology as a framework for others to use, whether (through) telephony or any other way of doing it," Osborne said.

    Sutherland praised the strategy, even if many solutions are a ways from reaching the market.

    "A large task"
    "IBM is more focused on incorporating voice recognition through the enterprise, whether it's going to be in (your) car or your desktop or your palmtop," he said. "IBM is taking on a large task. They have the capabilities to do it, and some of their competitors don't."

    As part of the strategy shift, Armonk, N.Y.-based IBM is introducing seven new products, all positioned for other companies to enable speech recognition in their products.

    While IBM is trying to make a big splash with the new products, many won't be ready until September or October--and some potentially later.

    Planned for autumn release is WebSphere Voice Server with ViaVoice Technology, a suite of tools for helping call centers better use the Web. The product supports VoiceXML and other emerging technologies, such as Wireless Application Protocol (WAP) and Voice over Internet Protocol (VoIP). The Voice Server software, initially available for Windows NT, will start around $15,000.

    Supporting its continued thrust into CNET's Linux CenterLinux, IBM later this month will begin selling ViaVoice Dictation for Linux, which will sell for about $60 retail.

    One of the more important products in IBM's broader voice-recognition strategy is Embedded ViaVoice, Multiplatform Edition. The software-development toolkit, which supports Java standards, is the backbone of a plan to get embedded voice-recognition technology into as many handhelds, cell phones and other wireless devices as possible.

    Osborne predicts that this kind of product could be a boon for companies developing Web devices for cars, as in the partnership between IBM and Motorola.

    But he admits that for now, drivers will have to keep reaching for the dial to tune in to their favorite radio stations instead of using voice commands.

    No guarantees
    "Depending on the device, it could be 12 to 18 months before you see a product. Particularly in the automotive industry, the product cycles tend to be long," Osborne said.

    "But the after-market sales for automotive are much shorter, so it depends on where you are and what you're looking for."

    Another product more immediately available will be CallPath Enterprise Foundation 6.3, a call center application that, when integrated with some Siebel Systems applications, integrates incoming telephone calls and Web transactions.

    The final three products to be unveiled tomorrow are versions of DirectTalk: Speech Recognition and Text-to-Speech for AIX, and Beans for Java.

    Despite its impressive tool set, there is no guarantee IBM can make its speech-recognition technology and tools the standard that everyone uses, Sutherland said.

    But in a market expected to reach $30 billion by 2006, that may not matter, Osborne said.

    "If you think of voice where the Internet was in 1994, you can see the kind of infrastructure IBM can assemble to take advantage of this explosion," he said.