Pattern Classification Frameworks?

[Steven Bethard]

Evan Klitzke wrote:
> What frameworks are there available for doing pattern classification?
> I'm generally interested in the problem of mapping some sort of input
> to one or more categories. For example, I want to be able to solve
> problems like taking text and applying one or more tags to it like
> "romance", "horror", "poetry", etc. This isn't really my research
> specialty, but my understanding is that Bayesian classifiers are
> generally used for problems like this.

In fact, a wide variety of classifiers are used in text classification, 
including Bayesian approaches, support vector machines, conditional 
random fields, etc.

> Are there any other frameworks I should be aware of?

I have used (but not recently) Orange:

     http://www.ailab.si/orange

I haven't used, but have been meaning to try, PyML:

     http://pyml.sourceforge.net/

A more recent addition (whose documentation needs work) is:

     http://montepython.sourceforge.net/

And here's a Summer of Code project to build an ML library:

     http://projects.scipy.org/scipy/scipy/wiki/MachineLearning

These are all general-purpose machine learning frameworks. So they can 
be applied to pretty much any classification problem (including the text 
classification problems you're looking at). You just need to pick out a 
set of relevant features to describe your data, and feed those features 
along with your chosen labels to a machine learning algorithm.

STeVe