The Bell Labs text-to-speech effort died on November 21, 2002, when the last of the active researchers on text-to-speech synthesis were laid off in the latest round of Lucent downsizing. Thus ended a half-century of research on, and investment in, this technology. The effort had been in danger for some time, first of all when Lucent Speech Solutions, the internal venture that was to be the outlet for Bell Labs' speech technology was closed down earlier in 2002, and second when the upper management at Bell Labs decided to eliminate all of speech research. The closing of LSS was particularly paradoxical: had it been kept it would have possibly been the only part of Lucent to break even this year, with a good chance of making a modest profit next year.

The death of TTS at Bell Labs is, of course, only a small part of a much greater tragedy -- the destruction of Bell Labs research. Some of the root causes of this outrage are well-known: the most obvious cause is Lucent's inability to support Bell Labs at anything like the level at which it had once been supported. This in turn is due to Lucent's current well-known financial woes, which (in turn) are due partly to the downturn in the telecommunications industry, and partly to Lucent's own corporate malfeasance during the late 90's including, but not limited to, outright fraud. But to some extent the woes of Lucent are a red herring: the seeds for the destruction of Bell Labs were planted at least in 1995 with the breakup of AT&T (many would say earlier in 1984, with the breakup of the old Bell System).

Lucent will fail. This is obvious since they are doing nothing whatsoever to address the fundamental problems that plague the company. One of the main problems is cronyism. Rich McGinn, who is surely the person with the largest share of responsibility for Lucent's current woes, walked away with a multi-million dollar severance package and continues to draw a one-million-dollar-a-year pension. Several top executives, it was recently reported in the Star Ledger, received large bonus packages within the last year even as the stock was plummeting: the company has argued that it needs to do this for "retention" as if they were talking about people worth retaining. Arun Netravali, who himself was involved in a transaction that smacked of insider trading and that netted him a few tens of millions of dollars, continues to occupy a nice office at Murray Hill and draw a handsome salary for "strategizing" about the future. At an even lower level, while all of the TTS researchers, and most of the speech researchers at Bell Labs are now gone, the managers are still there, though many of these have done little or no constructive work over the past five or more years. Lucent is thus clearly adopting the policy of maximizing the chaff-to-wheat ratio. If the Lucent upper management had any guts at all, they would immediately address these issues by cutting off McGinn's pension, demanding he pay back the majority of his severance package, fire top level executives who are drawing huge salaries for no work, and get rid of dysfunctional lower level managers. But Lucent's upper management has no such guts. They deserve to fail.

Unfortunately Bell Labs does not deserve to fail, yet it is probably too late to do anything about that. In 1996, at the time of the breakup of AT&T, one of our department heads came back enraged from a planning meeting on the future of Bell Labs, saying that "these assholes don't understand the value of the resource they are destroying." Unfortunately he was absolutely right, and in the intervening years they have never learned to use Bell Labs as anything more than a marketing symbol. Lucent has been the spoiled child that gets its first live pet, but fails to understand that unlike a stuffed animal, a live animal actually needs care.

The link to the Bell Labs multilingual TTS web page, which for many years had showcased the work, is here; but since no one is maintaining this site any more, it will probably not remain around much longer.

See here for a Polish translation.