Funny-Looking Numbers Cloud Speech-Recognition Realities
By Jim Seymour
Special to TheStreet.com
7/5/00 5:22 PM ET
thestreet.com
I hope you've been following Herb Greenberg's excellent reporting on the misadventures of speech-recognition vendor Lernout & Hauspie (LHSP:Nasdaq - news - boards) over the past couple of years -- and especially, over the past couple of months.
His column Wednesday noting the reporting curiosities from L&H lately left me laughing out loud. Read it for yourself, but Herb's observation that L&H's financials, taken at face value, show a spectacular boom in sales in Korea -- while falling everywhere else -- mocks a claim that's more than a bit hard to swallow.
In Herbie's words:
Where, then, did the big gain come from? None other than Korea. (Korea? Yes, Korea! The same Korea where tech sales, in general, are slowing!) Lernout's sales in Korea went from virtually zero ($97,000) a year ago to $58.9 million in the first quarter. That's almost triple the sales, during the same period, from either the U.S. or Europe. That's also more than half of Lernout's total sales in the quarter!
I don't know, maybe demand just exploded in Korea, and everyone there decided to start talking to their computers. (In English? Or are there Korean-language versions of L&H's products? Not clear.)
Or maybe L&H did a big bundling deal, with a fat upfront payment, with a Korean PC maker. Seems unlikely, too.
First, there aren't that many sizable PC makers in Korea, and none have talked about such a deal.
Second, upfront payments for software-bundling deals are never very big (at a likely bundling price of less than $5 per bundled copy, even a license for 1 million copies of L&H speech-recognition software would produce only a total payment, over time, of $5 million).
Third, a leap from $97,000 a year ago to $58.9 million in the first quarter of this fiscal year just boggles the mind. (Find me that salesman! I want to hire him for TheStreet.com!)
Anyway, I have no way to challenge L&H's reporting directly, and so I won't. But I do find it verrrrrrry... interesting.
Beyond financials, at some point, someone's gotta consider the quality of the products, too, in assessing the winners and losers. In my experience -- and I've been a heavy user of voice-recognition/dictation products on PCs for at least five years -- the Dragon Systems stuff has consistently been the best, with IBM's (IBM:NYSE - news - boards) Via Voice line second.
No voice-recognition product is perfect, but Dragon Systems' is well beyond the threshold of usability, which was a big step for the speech-recognition market. (For me, that's somewhere around 95% accuracy. That still produces a lot more shilly mistea$ks than I'd like, but the gain is worth the pain.)
Of course, now Lernout & Hauspie owns that line, too, after the sale of Dragon to L&H a while back. But watching the market, I don't see L&H replacing its own products or technologies with the Dragon stuff, but rather distributing both.
I think a big key to Lernout's future may be its ability to migrate toward the Dragon products, probably relabelled, and away from its own. Which, to be fair, aren't bad at all, but simply aren't up to the level, in my experience, of the Dragon Systems code.
What no one's talking about, though, is what's really driving the action in the speech-recognition marketplace, a force that will change the playing field for every participant.
Tomorrow I'll let you in on that little secret, which no one wants to talk about in public.
---------------------------------------------- |
