Don't LSA techniques usually involve finding the SVDhttp://math.nist.gov/javanumerics/jama/ seems to have just that. And from the NiST i would expect non-amateur software.
or eigenvectors of a large matrix? In this case I
would strongly doubt that something like this exists
in Java, although I've been wrong before...
Don't you generally do LSA with sparse matrices?I think you use singular value decomposition to make the matrix more sparse. I haven't used the NIST package, but it does supply an SVD class.
Otherwise they're gigantic because you can have
thousands of words. I'm not sure, though. But if so,
NiST doesn't seem useful.
coz lot matimaticalhttp://www.catb.org/~esr/faqs/smart-questions.html#writewell