Greg Kochanski

Classifiers for the Aesop Project

This is g_classifiers release 0.30.1, which was used in the Aesop project, more formally titled "Comparing dialects using statistical measures of rhythm" This was an ESRC research project to look at the rhythm of speech in different languages and dialects. See the project descriptions for more details.

The source code (and at Sourceforge ) implements linear and quadratic discriminant classifier forests, as described in the appendix to this paper. Technical documentation can be found here (in PDF format), or here (in HTML format).

(This above documentation is after the code was cleaned up a bit and documented. No functional changes were made. If you need the exact documentation, it can be extracted from the source code via epydoc.)

More recent versions of these libraries should be available from the Sourceforge speechresearch project. More recent documentation will be available either there or under http://kochanski.org/gpk/code/speechresearch/classifiers .

• This requires the gmisclib library to operate, which can also be found at Sourceforge, and is documented here, and
• the gpk_avg_py module to do some fast, robust, trimmed means. Also see the Sourceforge speechresearch project.
• It also requires Numpy, a python vector package,
• and Python 2.5 or 2.6, possibly later.