Will Furby be the breakthrough in speech recognition for toys?

New version of hit toy from Hasbro uses Sensory speech recognition and synthesis chip

Furby, a plush, electronically interactive toy from Hasbro, Inc.’s (NYSE: HAS) Tiger Electronics, was first introduced during the 1998 holiday season with great success (40 million pieces sold globally). On August 2, Hasbro announced that Furby has returned, with a sophisticated (for a toy) voice interactive personality, and is at stores now. The toy has an approximate retail price of $39.99.

The original Furby had no speech recognition and carried separate speech synthesis and microcontroller chips. The new toy uses the Sensory RSC-4128 chip from Sensory, Inc., with Sensory’s new FluentChip technology (SRU, July 2005, p. 7; May 2005, p. 6). The single chip provides not only speaker-independent speech recognition and speech synthesis (highly compressed recorded speech), but also acts as the “brain” (microcontroller) for the toy. Furby speaks English and Furbish, his “native” language. Versions that will be sold outside the U.S. beginning in September include Spanish, French, German, Italian, and Japanese. Todd Mozer, Sensory CEO, said that Hasbro was able to add Furbish by spelling the (invented) Furbish words using phonetic spellings using phonemes of the other language Furbish speakers. Sensory’s Quick T2SI tools support adding words to an application by spelling them as they sound.

When you a customer first buys a Furby, it speaks Furbish (and a little English).  The more one plays with him (using speech recognition, petting him, feeding him), the more English Furby uses in his vocabulary.  He replaces Furbish words with English words; for example, instead of saying “boo,” he’ll replace it with its English equivalent (“no,” for readers unfamiliar with Furbish). The interaction is designed to appear as if Furby is learning English.

Furby uses “Hey Furby” as a wake-up phrase. Furby responds with “Yeah,” “Huh,” or “What?” Then, the child (or adult!) can say one of a number of phrases (described in the FURBY instructions). The fact that Furby is not a native speaker of English gives the toy a perfect excuse for not understanding and asking for repetition.

Furby has motorized movement of its face and body, allowing it to display emotions. Furby also begins to spontaneously complain if he doesn’t get enough attention, requesting attention. Fortunately for parents, Hasbro has added an “off” button to this model.

Hasbro launched the toy with 40 young “delegates-in-training” assembled at One United Nations Plaza. A renowned linguist, John Henry Jorgensen, tutored the children on how to speak Furbish. Reportedly, by the end of the session, children who spoke different languages were talking to each other in Furbish.

Furby can dance, play a game, tell jokes, sing a favorite song, and talk to other Furby toys. Hasbro is differentiating individual toys by having various combinations of hair and eye color. Furby has a flexible beak, moving eyes, patented eyelid technology, moving eyebrows, moving plumage, moving and curving ears, and touch sensors on the belly and back.

Brian Goldner, Hasbro’s Toy Group president, said, “Technology enabled us to make the original Furby have personality. Today’s Furby is more vibrant and humorous because technology has advanced so much. Through cutting-edge innovation, Furby is even more magical than the original because it can actually hear and react to what a child is saying.”

Bill Meisel, Editor

Speech recognition has made significant progress, getting more accurate, less expensive, and easier to deploy, but there hasn’t been much recognition of that progress in the general business press or by the average person. Many people have been exposed to speech recognition in call centers when they phone customer service, but it hasn’t caused much of a stir. This lack of notice may be in part because speech recognition in call centers has been introduced incrementally, starting out with “enter or say your account number.” The next step was replicating touch-tone functionality and structure with speech, a practice I have criticized as “main-menu mentality,” rather than taking advantage of speech’s flexibility to help customers more easily achieve their objectives. More flexible and innovative applications have been deployed, but the perception seems to be that of evolution rather than revolution.

As Malcolm Gladwell wrote in his best-seller, The Tipping Point, there are often events that have a viral effect, making many people quickly recognize or adopt a trend. A successful toy (such as the previous article’s Furby), can make its way into millions of homes and gain the attention of the mainstream business and popular press more easily than a difficult-to-explain technology. The iconic image of the conversational but paranoid computer HAL in the movie 2001 makes speech recognition vaguely threatening, and may account in part for the tendency of the press to emphasize how today’s speech technology falls short of human abilities, a fact no expert contests.

On the other hand, a toy is friendly and non-threatening. Furby in particular is learning English, and any speech recognition failures can be attributed to this process. There is a distinct possibility that Furby (or some other toy or game in the future) could be the tipping point that makes the general public aware of the contribution that the Voice User Interface can make in many product and service categories.