Colin Beckingham is at it again with an new speech recognition article on IBM's developer works.  From the summary:

Summary:  A variety of plain-text, application-specific formats exists for the definition of non-natural language grammars for the preparation of a voice recognition model. Programmers can use the Speech Recognition Grammar Specification (SRGS) not only to express many of these formats in an open-standards structure, but also to define rules for the dialog manager necessary for interpretation of the output generated by the recognition model. Explore SRGS and Semantic Interpretation for Speech Recognition (SISR)-like methods using PHP in the context of non-natural language-specific grammars in this article.

--- (Edited on 11/9/2010 4:21 pm [GMT-0500] by kmaclean) ---