CMU Sphinx is Carnegie Mellon University’s (CMU) open source toolkit for speech recognition. The project has published this article on Sourceforge describing the automation of language model creation procedures as a complete C++ application. The article is the latest in a series posted on the Sourceforge page describing this ongoing open source academic research program.
For extensive documentation on this complex and fascinating project, visit the CMUSphinx wiki page.
My [Thai] wife has trouble understanding people with strong accents. Scottish and Indian accents are the worst. I’d like to see how well this stands up to accented users.
In one course, I had an electronics teacher who told us to use the “nouns”. We were all confused, until he told us to use them, to find the “un-nouns”. Then it finally clicked: knowns, unknowns.
> I’d like to see how well this stands up to accented users.
Easy, there is even a small tutorial on adaptation which you can use to raise the accuracy for accented speech:
http://cmusphinx.sourceforge.net/wiki/tutorialadapt