Real-time translation of Cisco Systems TelePresence virtual meetings, which a company executive said late last year would arrive in 2009, has proved much more difficult than the company expected.
The technology, which would combine speech recognition, a translation engine and text-to-speech conversion, was demonstrated at Cisco's C-Scape conference last December. Marthin De Beer, senior vice president of the company's Emerging Technologies Group, said Cisco expected the product to go on sale in the second half of this year with an initial set of 20 languages. Both Asian and Western languages would be included in that set, and users would have the option of viewing subtitles instead of hearing a digital voice, he said.
[ Stay ahead of the key tech business news with InfoWorld's Today's Headlines: First Look newsletter. ]
But at Cisco's big collaboration launch on Monday, where it unveiled 61 new products and features due over the next several months, the TelePresence translation system was nowhere to be found.
In fact, Cisco doesn't even have an estimate for when it will be available for any language, said Charles Stucki, vice president and general manager of the TelePresence Systems Business Unit, in an interview Tuesday. It turns out, Stucki said, that getting accurate translations was harder than Cisco's developers had expected. While they have had good results transcribing speech to text and rendering text as synthesized speech, automatically translating the written words has run into some hurdles.
"We haven't yet completed the engineering around exactly how we're going to implement the text-to-text translation," Stucki said. "The accuracy is not high enough that it's not a frustration for people."
To solve that problem, Cisco would have had to achieve a goal that has eluded some of the best minds in computing, according to Dan Miller, an analyst at Opus Research. In fact, the investment required would be probably too great for the company to ever earn back by selling a product, he said.
"There's every reason to believe that government projects are pretty far along in certain selected languages ... but it's going to be a while before there would be an economically viable 20-language simultaneous translation component to TelePresence," Miller said.
He predicted a fully automated, real-time system wouldn't come for at least three years, and maybe never.
"There's some human interaction to make it accurate. I think that's always going to be true," Miller said. "Any (translation) application anyone develops, there's got to be some kind of review cycle."
But analysts applauded the progress Cisco has made in transcribing speech to text in English. Cisco's Media Experience Engine (MXE) 3500 and 5600, announced Monday and coming in the first half of next year, will be able to convert speech to text nearly in real time when users record videos with the Cisco Show and Share application. Show and Share, also announced Monday, is a system for enterprise employees to create video messages for their co-workers.
The speech-to-text technology only works in English today, and Cisco's Stucki said it's not intended to form the definitive document of a video or meeting. Instead, it's designed for searching the content for topics of interest and going to that specific section of the video, to save time. The capability is part of a broader initiative by Cisco to make all of an enterprise's content searchable for key subjects, which will also make it possible to organize all types of documents and communications by topic.
"That'll have an impact on productivity and forming of groups to work on projects ... in ways that wouldn't have happened before," Opus's Miller said.
In trying to make all information in an organization available to everyone who needs it, immediately, Cisco may have simply taken on a bigger challenge than it could readily handle, analysts said.
"On this, I would give them a break. It's very tough to do," said Gartner analyst Ken Dulaney.
Meanwhile, Cisco may not lose too much face for not delivering on such a futuristic vision, according to Abner Germanow, an analyst at IDC.
"I don't know that there are many people who are really waiting for it," Germanow said.