Once I tried to organize ~100 documents I collected through keyword search in IEEE. And as I couldn’t find a way to tag pdf files, I implemented one in php. Then I tried to classify them through multiple taxonomies at the same time, each being multilayer/singlelayer, texture/color etc. However, it required manual tagging of every doc and the complexity made it unmanageable. Eventually, I gave up tagging and returned to hierarchical organization.
I think tagging is not a suitable way to manually-artificially classify content, because it is similar to a semilattice organization. Also, no folksonomy will naturally yield a hierarchical tree structure. The following article explains this distinction in the context of architecture:
“A City is Not A Tree”
by Christopher Alexander
artificial organization: tree => hierarchy
natural organization: semilattice => tagging
If you are interested, I recommend you this book.