Sidebar: Origins of Speech Synthesis

Computerworld - We've paid a lot of attention over the years to speech recognition -- getting the computer to hear and understand what we say to it -- but much less to how the computer talks back to us via text-to-speech (TTS) synthesis. A number of proprietary systems have been used, beginning in 1961 when Bell Labs researchers programmed an IBM 7094 to sing "Daisy" (an event memorialized in the film 2001: A Space Odyssey).
TTS got a big boost in 1976 when Ray Kurzweil introduced his Reading Machine, which could scan printed copy and speak it. At the time, we acknowledged the significance of the breakthrough but made bad jokes about the heavy "Swedish chef" accent.
In the intervening three decades, computers have grown far more powerful and we have developed many more tools with which to produce better output.
VoiceXML has its roots in an AT&T Bell Labs project called PhoneWeb, a phone markup language and platform for building both consumer and call-center telephony applications.
Motorola used a similar approach so it could provide mobile users with interactivity and current information. Motorola focused on hands-free access, emphasizing speech recognition rather than touch-tones for input. Motorola based its language on XML and in 1998 announced VoxML.
Another XML-based speech language, Sable, was also introduced in 1998 and built on the Spoken Text Markup Language (1997's STML, based on the 1996 Speech Synthesis Markup Language) and Java Speech Markup Language, also dating from 1997.

Read more about Enterprise Architecture and SOA in Computerworld's Enterprise Architecture and SOA Topic Center.