Today, The New York Times reported that Amazon’s Kindle 2 much vaunted text-to-speech (TTS) capabilities, provided by Nuance Communications, came up short when trying to pronounce President Barack Obama’s name. The device uttered something closer to Baah-raah-k O-baah-maah (closer to the sounds in “black” and “Alabama,” the Times said. The paper adds that the problem has since been corrected. Obama’s name has added to the Kindle’s TTS dictionary and will be included in the next wireless update.
The Kindle TTS misfire came to prominence as many news organizations began openly speculating on whether subsequent versions of the Kindle could create a viable non-paper-based means of distribution. Wired, for instance was running the headline “How the Next Kindle Could Save the Newspaper Business” in stories about partnerships the The New York Times and Washington Post were looking to hatch, while mediabistro.com pondered, “Can The Kindle Save Newspapers?” Whether any of that’s true, the failure of Kindle’s TTS to pronounce things like the President’s name correctly may put at least a temporary crimp in any role speech might in any Kindle paper-saving venture.
When it comes to that though, don’t blame Nuance.
According to Mike Thompson, senior vice president at Nuance Mobile, the company invests “significantly in the technology of audio output.” Nuance is committed to constantly updating its grammars and dictionaries with data gathered from its Dragon products.
Pointing to Nuance’s street name domain, its business name domain, and its email and SMS domains, he says, “We invest and pour money into the continuous optimization of that over time. Over time, things change and there’s continued investment and refinement.”
“It’s always an ongoing exercise. It’s a little bit like when someone new jumps on to the scene and the web search companies haven’t had many searches for them. They have to optimize around that,” he adds.
Thompson claims that Obama’s name was added to Nuance’s TTS dictionaries almost 12 months ago, which at least on face value seems to suggest that Amazon’s version is more than a year old.
Thompson, for his part, could not comment on which build of Nuance’s TTS engine was packaged into the Kindle 2.
If the engine is more than a year old, however, that fact alone may present real problems to users who would want to user TTS capabilities to read the news. If familiar names and places in the headlines are rendered into incomprehensible pronunciations who’s going to want to listen? To deliver accurate pronunciation though, Amazon would have to commit to flashing its engine with updates to account for the names of new w0rds and names that come into play, a significant undertaking by any estimation
Considering Amazon’s recent troubles with the Authors’ Guild complaint that the TTS was a violation of their exclusive books on tape rights, the company may not see putting additional weight behind the functionality as a winning proposition.
On that count, Nuance would not comment, but to say that the market for TTS is strong as the quality of the technology is “excellent” and that many domains are embracing it.
Amazon itself could not be reached for comment.
