We’re still creating the language model portion of Sphincs(sphinx4), but at a minimum, Sphincs(sphinx4) will support three type of language models. The simplest type of language model is no language model at all, and is used in the case of isolated word recognition. The next common language model type is a context free grammar that can used in speech applications based on command and control. The final language model type is based on n−gram grammars, and is typically used for free form speech.
Since we’re still working on the format of the language model, this section is expected to change over time.

Last edited Oct 12, 2014 at 5:23 PM by DxN, version 2