SpeechWeb

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

A SpeechWeb is a collection of voice-controlled web applications you can access with speech-enabled browsers on your device. You navigate by speaking commands, and links open the next speech page.

The idea goes back to 1995, when Hemphill and Thrift developed a system where pages were downloaded to the user’s computer and opened by speaking to activate links. In the mid-1990s, researchers at AT&T talked about a new way to let people use the web over regular phones. Between 1995 and 1999, AT&T, Lucent, Motorola, and IBM created their own speech and phone markup ideas, forming the VoiceXML Forum and designing the Voice Markup Language (VXML), which the W3C adopted in 2000. VXML is used to build hyperlinked speech applications. A VXML page can prompt user input, run recognition grammars, output synthesized voice, control flow, call JavaScript, and link to other VXML pages downloaded as you browse, similarly to linking pages on the traditional web.

Around the same time, a research group at the University of Windsor in Canada explored another approach. They built client-side speech browsers that download a specific recognition grammar from each speech application site. The browser recognizes spoken input locally and sends it to a remote server, which processes it and sends back text to be spoken aloud. The term SpeechWeb was used in 1999 to describe this network of hyperlinked speech apps. The first SpeechWeb browser was shown at the AAAI conference.

The word speechWeb has also been used in other contexts, such as speech‑language pathology, and the term was trademarked by a company that later became part of HP. By 2005, few public voice applications existed, even though VXML was mature, and most apps were built for business use. This led to the idea of a Public-Domain SpeechWeb—accessible through ordinary web browsers with speech plugins and created by the public, much like regular web pages. A public-domain SpeechWeb browser was demonstrated at the 2007 World Wide Web Conference in Banff, Canada. It was a small X+V page run by the Opera browser with an IBM speech-recognition plugin. Two research groups are developing software to help non-experts build and deploy SpeechWeb applications.