L&H Assets Auctioned: Company To Cease Operations By Year-End

11/28/01

We’ve been largely silent on the L&H “situation,” on the theory that the Wall Street Journal and the New York Times were reporting all the juicy parts and that, as a former consultant to the company (not on financial strategies, need I say), discretion might be in order.  However, now that things are winding down, it seems like a good time to comment on how the market valued L&H’s distressed assets and on the voice processing market more generally.

For those of you, who have little interest in Voice Processing, just skip this article.  I’ve been following this technology for 26 years, so I feel like I should provide some sense of history here.

Voice technology wandered in the wilderness for a long time.  We could recognize speech but you had to give up one of three things:  accuracy, speed, or reasonable cost to get the job done, so we were mainly looking at science projects, not products, until personal computers became powerful enough, in the mid-90’s, to offer a reasonably low-cost and powerful platform.  That permitted many users to try out speech, usually as a PC interface or for dictating text into word processors.  It caught on as a novelty item, but it isn’t yet a mainstream product.  Users’ expectations continue to exceed product capabilities (which are enormously higher now than when I saw my first 10 word system in 1975).

In the meantime, we learned that you could usefully use voice recognition, in more defined applications (called narrow domains in the AI vocabulary on which they are based) quite successfully.  Voice came into wide usage for server-based applications like telephone directories and customer service.  With increased proficiency in getting computers to “read” digital files in more human voices, Text-to-Speech also became a burgeoning business for applications like reading a driver directions as he drove his car or, less exotically, reading email to someone with limited sight.

The market was full of tiny firms, but very large companies had done much of the early research and continued to be players, especially IBM.

As the market heated up, a tiny Belgium company, L&H (Lernout and Hauspie) opened offices in the U.S., invented and acquired an expanding portfolio of technologies, and in the Big Bull market of the late 90’s convinced investors that the company was worth, at its peak, some $9 billion.  Unfortunately, much of that value seems to have been inflated by fraudulently reported income and a scheme to convert R&D expenses to revenue.  When the Wall Street Journal uncovered the edges of that fraud and pulled on the string, L&H came apart, resulting in its filing for Chapter 11 bankruptcy in the U.S. and Belgian courts last November (2000).

After a variety of tries to keep the company afloat, a decision by the Belgian Court forced L&H to sell off its assets.  A number of pieces of the company were sold this fall and an auction was held last week, with the results announced today.

Most of the speech processing assets went to OCR software vendor ScanSoft of Peabody, Massachusetts.  If that sounds bizarre to you, we’d remind you that both OCR scanning and voice recognition are based on AI pattern recognition, so there’s some underlying commonality.  Also, both companies were in a business where they sold software engines to partners and packaged software through retail channels.  Paul Ricci, ScanSoft’s CEO and Chairman notes that the applications are also complimentary, both focused on improving productivity.  “There are large operational synergies,” he noted, ranging from fully leveraging operational infrastructure, similarities in distribution, horizontal expansion into new market opportunities, and economies from similarities in vertical market interests (such as medical, legal, financial services, and public safety).  ScanSoft bought L&H’s Text-to-Speech and some ASR assets (including Dragon’s Naturally Speaking product line) for $39.5 million in cash ($10M), stock ($26M) and a note ($3.5M). 

Also in the auction, L&H sold off an additional $6.75 million in assets including the ASR technologies of L&H's ISI division to Pennsylvania-based Multimodal Technologies ($2M in cash and the assumption of a $2M earnout); L&H's Intelligent Content Management (ICM) assets and its Knexys division, which develops next-generation ICM technologies by Pennsylvania-based Vantage Technology Holding )$2M); L&H’s Audiomining assets (audio search engine technology pioneered by Dragon Systems Inc.) by Dragon Catalyst LLC, a newly formed company headed by Dr. James Baker (0.75M).

Prior to the auction, L&H had already sold its Mendez translation unit to Bowne & Co. ($44.5M ); its medical transcription unit to MedQuist ($25M); its Kurzweil Educational Systems Group to executives of that group ($2M)’ and its Application Technology (computer translation) to  Apptek Partners ($2M) for a total of $75.5M.

With the auction sales, which will close in December, that will value the asset sales at $113.75 million, not much against the once flamboyant market cap of $9 billion (or against the secured debt of over $400 million).

Caught in the debris of this chaos were two well known speech processing firms, Dragon Systems and Dictaphone Corporation, both acquired in 2000 by L&H for hundreds of millions of dollars of its now essentially worthless stock. 

It’s hard to feel sorry for the corporate denizens of Dictaphone (who subsequently managed to separate the corporation from the L&H bankruptcy and file for its own Chapter 11 reorganization, as well as to pay for the licensing of certain L&H assets and to employ some L&H speech specialists).  An October 27 letter on Dictaphone’s website points out that the company is nearly out of its bankruptcy process and on the way back to profitability.

Dragon is another story.  This firm, started by two former IBM research scientists, Jim and Janet Baker, had created the best-rated PC-based voice recognition product.  To have their years of work worth nothing and their loyal employees facing unemployment was excruciating.   The Bakers had hoped to be a successful bidder in the L&H asset auction, but the complex process, which favored bids for multiple assets, played against them.  Jim Baker’s company did manage to buy one Dragon asset, the audiomining technology Dragon pioneered, for $0.75 million.

Paul Ricci of ScanSoft notes that throughout the software industry consolidation is the order of the day.  L&H’s demise and the fate of its assets fit into that pattern.  It also anticipates further departures.

Speech processing is getting more mainstream each year.  Some applications, such as telephony-based operators interacting with callers in natural language have become quite common.  Voice interfaces for PC’s are doable, but still not quite ready for prime time.  Wireless, handheld devices will demand, however, voice interfaces for users who find their small size, required for portability, makes it difficult to use a keyboard or pointing device as the interface.  At the same time, we probably don’t need hundreds of such vendors, particularly in an area where the technology is difficult and expensive to produce and requires scarce, specialized skills.  We’d expect consolidation to continue – and that hints of further purchases coming for both Mr. Ricci at ScanSoft and his fellow survivors. 

In the meantime, users can rest easy.  If the investors are disappointed with the L&H outcome, the users’ interests of continuing, supported product in the marketplace will be well served.  That’s too important for users to continue experimenting with new technology for investors to ignore.  Maybe they should just write this one off and look for the next good investment.