10.1 Speech Recognition Grammar Format

The Speech Recognition Grammar Format (SRGF) is an XML language for writing voice recognition grammars. It is fully embedded within VoiceXML. Historically, SRGF derives from JSGF, the Java Speech Grammar Format.

As a historical note, the VoiceXML 1.0 specification made extensive use of JSGF in examples. However, JSGF was never required or even recommended for VoiceXML 1.0 browsers.

VoiceXML 2.0 compliant browsers must support SRGF. Other grammar formats may also be supported, depending on the browser.

There is an alternative form of SRGF which is not based on XML but on Augmented Backus-Naur Form (ABNF). Backus-Naur Form (and its Augmented and Extended brethren) are familiar to computer scientists as a way of representing programming language syntax.

The ABNF form of SRGF has the advantage that it may be easier for people to read and write. It also has the advantage that it may be possible (depending on browser support) to specify ABNF grammars in-line in a VoiceXML page, unlike the XML form which must always be in a separate file (at least, according to the current draft standard, this is likely to change). The capabilities of the two flavors are equivalent, and they are designed so that automatic translation between the two formats is straightforward.

Browsers may also support other grammar formats. The only format which is mandatory for a VoiceXML 2.0 compliant browser is the XML form of SRGF.

Proposed MIME types are "application/grammar+xml" for the XML form and "application/grammar" for the ABNF form.