CeBIT: Speech technologies come of age

Mainstream acceptance of speech technologies originally designed for people with special needs will lead to 'explosive' growth within the next couple of years, IT executives and vendors increasingly agree. An array of products that use technologies like natural language understanding, speech recognition and text-to-speech on PCs and for telephony are being demonstrated at the CeBit trade show.

Mainstream acceptance of speech technologies originally designed for people with special needs will lead to "explosive" growth within the next couple of years, IT executives and vendors said repeatedly here this week.

An array of products that use technologies like natural language understanding, speech recognition and text-to-speech on PCs and for telephony are being demonstrated at the CeBit trade show by a variety of companies, including Lernout & Hauspie Speech Products Inc. (L&H), IBM Corp., Dragon Systems Inc., Philips Electronics NV, Mannesmann Arcor AG & Co. and CyberLab Interactive Productions GmbH.

Companies plan to release products in more languages throughout this year and expect continued improvements in the technologies. These will lead to improved accuracy both in applications that have to understand speech and those that translate languages, as well as continued advances in overall quality.

L&H demonstrated a multilingual chat application here that converges a number of speech-related technologies and allows people to communicate in their natural language to others who might not speak the same language. In the demonstration, one person spoke in English and his words were translated to a German-speaking woman in her native tongue, and vice versa.

The company intends to hold off releasing the software until later this year because "the voice (that speaks the translated words) sounds like a robot. We want to introduce it with a real human voice," said L&H President and Chief Executive Officer Gaston Bastiaens after the demonstration.

While he forecast that 2001 will be a key year for the explosion in market growth he anticipates, Bastiaens also noted that there are predictions that by 2015 the computational power of microprocessors will exceed the power of the human brain, further boosting speech technologies and applications. By then, however, it seems likely that those products will be firmly entrenched in the mainstream with smart phones serving as the "dominant piece" in pushing along the market.

Also critical is the weight of companies like IBM, which has developed technologies and gotten products into the market, and Microsoft Corp., which has licensed technology from L&H and intends to use speech-recognition in a host of its future software releases, including a new version of Office that will be available in 36 languages.

"They are not going to develop for all those 36 languages themselves," Bastiaens said, "so we will be there."

L&H, based in Ieper, Belgium, has created its own research center to work on technology, including hoped-for breakthroughs in artificial intelligence. While some companies choose to contribute to research universities and institutions, Bastiaens said that "the only thing you're sure of after awhile is that the money is gone." The L&H approach has meant that the company has attracted university researchers and others to come work at the center, he said.

L&H is among the companies announcing expanded language offerings during the trade show. Likewise, Dragon Systems, based in Newton, Massachusetts, announced here this week that its Dragon NaturallySpeaking Mobile and NaturallySpeaking Legal Edition will be released in German.

Besides the expanding need for additional language capabilities, growth in Internet use also is helping the market for speech technologies. IBM recently announced a talking World Wide Web browser and the German firms Mannesmann Arcor and CyberLab Interactive Productions announced a system today at CeBit that enables Internet access over the phone without a computer. Their TalkingWeb voice browser was created by both companies and lets users call a phone number and log online with voice commands. The browser reads Web page content to the users and also allows users to participate in chat sessions and play and record audio files.

IBM also has emphasised its telephony offerings here this week, demonstrating various products and technologies using natural-language understanding. The company is expanding its offerings to German speakers and also plans to introduce a telephony application in U.K. English before year's end. Additional languages are in the works, said Edward Zinnes, North American marketing manager for speech-enabled technologies.

The natural-language understanding products and technologies from IBM can be used for a range of telephony functions. Users can access call centers, banking and financial services and the like and IBM also is exploring applications for use in the travel industry.

"Of course, we are expecting an explosion in the market, but there are some prerequisites," said Wolfgang Karpstein, IBM marketing and sales manager for all speech products in Germany, Austria and Switzerland.

Chief among those prerequisites is the ability to "talk to the device as you would talk to somebody," he said.

That means using full sentences and not just key words or phrases. It also means talking to an application that will remember what has been said and not require needless repetition.

Vendors are convinced that the speech-technology market is going to take off because it affords computer and telephone users all manner of conveniences, from hands-free PC use, to ordering the VCR to record a favorite show, to obtaining driving directions via the Web while in the car using a voice-activated computer in the vehicle.

"It puts the user back in control of the application," Zinnes said.

L&H, in Ierper, Belgium, can be reached on the Web at http://www.lhs.com/. IBM, in Armonk, New York, can be reached at +1-914-765-1900 or at http://www.ibm.com/. Dragon Systems, in Newton, Massachusetts, can be reached at +1-617-965-5200 or at http://www.dragonsystems.com/. Mannesman Arcor, in Eschborn, Germany, can be reached at +49-069-2169-3212 or at http://www.arcor.net/. Cyberlab, in Germany, can be reached at +49-089-9250-2354 or at http://www.cyberlab.de/.

Join the newsletter!

Error: Please check your email address.
Show Comments

Market Place

[]