Language Factory

A fully configured speech decoder uses a set of different models, including acoustic, lexical and language. The tools on this page will allow you to easily construct lexical and language models consistent with the formats in use in the ARPA speech community (and by others). They are part of the Sphinx Knowledge Tools.

Web-based Tools

A statistical language model compiler suitable for small corpora.
Creates a Sphinx-compatible pronunciation for any word you give it (accuracy not guaranteed!)
Transforms text into a plausible spoken equivalent. (that is, sounds out numbers, deals with abbreviations, etc.)


The code for the Quick_LM language modeling tool has been released into Open Source.[12 June 2002]
Alex Rudnicky
Last modified: Wed Jun 12 17:47:49 EDT 2002