I've been working on a new project I call “Japanese Dependency Vectors” or “ jpdv” for short. It's a program that generates dependency based semantic vector spaces for Japanese text. (There's already an excellent tool for doing this with English, which was written by Sebastian Pado.)
However, jpdv still has a way to go before it works as promised. So far the tool can parse CaboCha formatted XML and produce both a word co-occurrence based vector space and a slightly modified XML representation that better demonstrates the dependency relationships of the words in the text.