Microsoft’s Acquisition of Nuance Communications: A Second Go at Language Processing

·6 min read

Welcome to the Capital Note, a newsletter about business, finance, and economics. On the menu today: the history of Nuance’s Dragon brand, J&J vaccine halted, Jack Ma capitulates, and W. Brian Arthur on “Economics in Nouns and Verbs.” To sign up for the Capital Note, follow this link.

Dragon 2.0
Yesterday, Microsoft announced a $16 billion acquisition of language-processing firm Nuance Communications, the latest in a string of big-ticket deals closed by CEO Satya Nadella, who spearheaded the purchases of LinkedIn and Github, among others.

With the addition of Nuance’s cutting-edge language-processing technology (the lion’s share of which is in the health-care sector), Microsoft hopes to beef up its enterprise cloud offering. Nuance’s Dragon Ambient eXperience (DAX), released last year, has gained widespread traction among physicians and nurses for clinical documentation, allowing doctors to legibly dictate their notes.

While Nadella cast the deal as transformative, it’s not the first time Microsoft has bet on Dragon–brand speech-processing technology. Decades ago, Microsoft lost millions on an investment in Lernout & Hauspie Speech Products (L&H), then the corporate owner of an early version of Nuance’s speech-processing technology. Founded in 1987, L&H grew to a valuation of nearly $10 billion by 2000 through aggressive M&A, including acquisitions of Berkeley Speech Technologies, Globalink, Dictaphone, and, in mid 2000, Dragon Systems — the birthplace of Nuance’s Dragon brand. Shortly thereafter, at the height of the Dotcom bubble, the company was revealed as a fraud:

[L&H] improperly reported revenue from barter deals with other software firms in which no cash changed hands; immediately recognized revenue for sales that were contingent on L&H later performing development work for the customer; and sometimes reported sales before contracts were signed, when it was unclear the customer had the ability to pay or when the customer’s ability to pay depended on investment from L&H.

At the time of L&H’s bankruptcy, Microsoft owned 5 percent of the company’s shares, with Intel and Michael Dell also left holding the bag. James and Janet Baker, the founders of Dragon Systems, lost most of their net worth after accepting L&H’s all-stock offer:

The corporate takeover of Dragon Systems in an all-stock deal went terribly wrong. [The founders] lost everything when Lernout & Hauspie was revealed to be a spectacular fraud. L.& H. had been founded by Jo Lernout and Pol Hauspie, who had once been hailed as stars of the 1990s tech boom.

They are widely credited with advancing speech technology far faster than anyone thought possible, primarily because of an epiphany Mr. Baker had while doing his doctorate research. He figured out that speech recognition could, in essence, be reduced to math. You didn’t have to teach a computer to recognize accents or dialects, Mr. Baker realized — you just had to calculate the mathematical probability of one sound following another. His algorithms proved remarkably accurate and eventually became the industry standard. (Want to know more? Ask Siri.)

The Bakers founded Dragon Systems in 1982 in an old Victorian house in West Newton, Mass. At that time, despite having two school-age children and a big mortgage, they were determined to take no venture capital and to finance the company’s growth with its own revenue — once they had a product. They figured they could last 18 months, maybe 24.

Soon thereafter, the Bakers created DragonDictate, a speech-to-text product akin to a primitive version of Apple’s Siri. After more than 15 years developing software, Dragon needed more capital and accepted an all-stock merger offer from the aforementioned L&H.

The botched sale was a live issue as recently as 2013, when the Bakers took their Goldman Sachs bankers to court, alleging that the dealmakers had not conducted sufficient due diligence on the company. Goldman won the suit at trial.

L&H would end up being purchased by ScanSoft, which merged with Nuance Communications in 2005. Fast-forward 16 years, and the Dragon brand — many times recycled and remade — has ended up back in the hands of Microsoft. Such is the long arc of technological progress.

Around the Web
FDA halts use of Johnson & Johnson vaccine

Federal agencies will call to temporarily suspend the use of Johnson & Johnson coronavirus vaccines, after six recipients of the vaccine developed a blood clotting disorder.

Almost seven million Americans have received the Johnson & Johnson vaccine, which is administered in a single dose. Six recipients, all of them women between ages 18 and 48, developed an extremely rare clotting disorder, known as cerebral venous thrombosis.

Jack Ma’s Ant Group converted to financial-holding company after regulatory crackdown

In a statement, the People’s Bank of China said Ant representatives were summoned to a meeting Monday with four regulatory agencies that also included the country’s banking, securities and foreign-exchange overseers. It said a “comprehensive, viable rectification plan” for Ant has been formulated under the regulators’ supervision over the past few months.

The directive follows an intense regulatory assault on Mr. Ma’s business empire that began with the suspension of the company’s blockbuster initial public offering in November. Ant had been on track to sell more than $34 billion worth of stock and list on stock exchanges in Hong Kong and Shanghai, when Beijing pulled the plug on the deal after Mr. Ma criticized financial regulators in a public speech.

John Cochrane on today’s CPI release

To the extent that we’re just seeing “reflation,” a return of prices to normal after a steep covid-induced recession, the graph suggests that was over last summer. “Reflation” was over by September. . . . Reflation was over last fall and this is something new.

Random Walk
Economist W. Brian Arthur recently published a paper called “Economics in Nouns and Verbs” arguing for more agent-based economic modeling. It’s worth reading in full, but here’s an excerpt:

Let me begin by pointing out that economics deals with prices, quantities produced, consumption, rates of interest, rates of exchange, rates of inflation, unemployment levels, trade surpluses, GDP, financial assets, Gini coefficients. These are all nouns. In fact, they are all quantifiable nouns — amounts of things, levels of things, rates of things. Economics as it is formally expressed is about amounts and levels and rates, and little else. This statement seems simple, trite almost, but it is only obvious when you pause to notice it. Nouns are the water economics swims in.

Of course in the real economy there are actions. Investors, producers, banks, and consumers act and interact incessantly. They trade, explore, forecast, buy, sell, ponder, adapt, invent, bring new products into being, start companies. And these of course are actions—verbs. Parts of economics—economic history, or business reporting—do deal with actions. But in formal discourse about the economy, in the theory we learn and the models we create and the statistics we report, we deal not with verbs but with nouns . . .

So why then is economics noun-based? One good reason is that any field, be it musicology or jurisprudence or architecture, as it becomes theoretical, tends to categorize and marshal its thinking into concepts and often these concepts are nouns. They are real or abstract things. But this can only be a partial explanation. Concepts don’t need to be nouns, they can also be verbs—processes—as is true in biology, so something else must be at work.

That something else, I believe, is that formal economics — theoretical economics — is expressed very largely in algebraic mathematics, a language that allows quantitative nouns only, but doesn’t allow actions.

— D.T.

To sign up for the Capital Note, follow this link.

More from National Review