Matt Hodgson has three posts on taxonomies, folk taxonomies and Folksonomies.
I have struggled with trying incorporate taxonomies into our work product retrieval and precedent storage systems. It has been difficult to create anything that exists beyond a single practice area. As Matt highlights in his post, taxonomies create a one-size-fits-all approach that may be comprehensive, but it is hard to get all of the concepts in place. As the taxonomy gets bigger it just makes it harder for the user to find topics and harder to file items into the taxonomy.
We have implemented a tool that auto-categorizes documents: West KM. It does an excellent job of categorizing substantive legal memos into the West key system. But is it worth the effort? In our analysis, users start with search terms. Very few browse through the categories. I also find the west taxonomy to be lacking in depth in many areas.
Folksonomies are intriguing to me, because they put categorization in the users hands (for better or worse). I have become a big fan of del.icio.us (you can see my tags).
Folksonomies have a lot of flaws. Just take a look at the most popular tags on Del.icio.us. "video" and "videos" are two separate entries.
What really intrigues me about folksonomies is being able to include them as part of the metadata for an enterprise search. As we have experienced with WestKM, most users will start with a search. If the tags allow us to manipulate the rankings of documents it could make the enterprise search engine more powerful and useful.
I assume the enterprise search engine will use stemming and other techniques to eliminate the noise of the "video" and "videos" tags.