January 02, 2005

Please evacuate the area: a speech synthesis story

While I stood in line for a slice of pizza at Sbarros opposite Gate 24 on concourse B of the Philadelphia airport, waiting to board my flight back from that great city yesterday, a loud alarm went off at Gate 23. A constant, strident, teeth-jarring, metallic scream tone started up (eeeeeeeeeee!). Overlaid on it was a sequence of rising alarm tones ("Whooeee! Whooeee!!") very much like the great call of a female gibbon. Then, added to these two earsplitting sounds, came a harsh, synthesized (or could it have been prerecorded?) male voice saying, with the familiar failure to destress grammatical items such as articles like the and auxiliary verbs like have and the official insistence on technical words like activate rather than set off: "Thee fire alarm in this area hazz been activated! Please move away and evacuate thee area!" (People were not eager to miss boarding their 100% full flight, so they stood up slowly and uncertainly, as if very reluctantly participating in a standing ovation for a someone who had never really been all that popular. The Sbarros staff stubbornly kept on handing out slices with pepperoni and mushroom: no sense in losing business just because the concourse is about to be engulfed in flames.)

After five minutes of this aural purgatory, humans intervened, but the result was just that a live female voice on the public address system was added over the top: "The fire alarm is a false alarm. Someone opened a security door. Please do not evacuate the area." The male voice (not being human) refused to be put off: "Thee fire alarm in this area hazz been activated! Please move away and evacuate thee area!" So the woman began to contradict the voice more emphatically, over the scream tone (eeeeeeeeeee!) and the gibbon call ("Whooeee! Whooeee!!") and the male voice ("hazz been activated!"): she insisted, "Repeat, the fire alarm is a false alarm! Please do not evacuate the area!"

It was a quarter of an hour in absurdist, synthetic audio hell. The people like Douglas "Hitchhiker's Guide" Adams who have suggested that our technology (particularly the technology incorporating mechanical speech) is annoying if not malign (remember the doors that said "Glad to be of service!" every time they opened?) are not wrong. They have barely scratched the surface.

Posted by Geoffrey K. Pullum at January 2, 2005 12:57 AM