I’m not a big fan of "tagging" although it’s popular in the so-called web 2.0. Reminds me too much of old-fashioned labeling and filing. I keep hoping smart computers will do the work for me. Here’s an article about a computer system that generates ontologies for language data (as opposed to "folksonomies" in the tagging world). It uses several sources of input to develop a more robust way of analyzing text than you get with, say, Google.
I live in hope.