So, I thought I would see if I could crank out a simple converter. Does anyone else have such a need?
I was looking into running the Reuters tests with Mahout, just to see
how things go, and, being efficient (lazy?) I wanted to avoid having
to do all the "pre" work if I could, so I found that Weka has an ARFF
formatted set of the files (http://sourceforge.net/projects/weka/
files/) for the ModApte tests. In looking at the format and http://www.cs.waikato.ac.nz/~ml/weka/arff.html
it seems that this corresponds fairly well to our vectors (which
makes sense) in that the attributes at the top are our Vector labels,
and, of course, the data section corresponds to the data.
