Actually, several wikis allow the creating of sibling pages within a "namespace" so that you can have WikiLinks like ProgrammingLanguages/Java, ProgrammingLanguages/Python, etc. You can then go to ProgrammingLanguages and see links to ./Java and ./Python. The Apache Wiki supports this, for example. (I'm not sure what the underlying software is.)

SnipSnap has also added namespaces for entries (aka "snips") for the next release. They have also just added per-snip labels which could probably help with "refinements". (I don't know too much about how that works in SnipSnaps, nor am I sure I know what you mean by "refinements")

SnipSnap looks pretty cool. Their "labels" seem to be a mechanism for metadata tagging, which is a prerequisite for refinements, but not sufficient in itself.

The namespace idea you mention looks like it would restrict the ProgrammingLanguages/Java page to live only within ProgrammingLanguages -- you couldn't also put it under People/JamesGosling/Java without creating two different "Java" pages. So that's not really what I'm looking for. The whole point of faceted classification is that things can be classified in multiple overlapping hierarchies.

I've installed SnipSnap on my work machine and I'm currently using it to organize my brain. (I hate it when I get distracted and then forget what I was doing. Hopefully SnipSnap will help.) It has quite a few good ideas.

"First, wikis don't suggest refinements (other categories which could be intersected with the current category in order to narrow the search). Second, wikis don't do hierarchical categories. The combination of both of these features enables a qualitatively different kind of site."

Kim,

The second part isn't true, you can backlink a category to other categories - may not be as rich as some would like.

For the first part, can you be more specific as to what you're after? If I understood it better, I might be able to implement it.

True, wikis aren't as rich say as a Topic Map or RDF, but I think they do go a long way giving the minimal effort involved (a lot of the problem wiht KM stuff is that it burdens the user upfront).

One of the problems I see with multi-faceted classification is the proliferation of classifications. My own blogging software allows for the classification of each entry; in fact, it allows multiple classifications for each entry (and although every entry of mine has such classifications, there isn't support yet for using the data for anything).

Out of 1,514 current entries (nearly four years worth) I have created 2,671 distinct classifications---1.76 catagories per entry). Part of that is unintentional duplication (five different typings of “Cue Cat” for example); silly classifications are another reason (“dead zombie langauges rising from the dead” anyone?). It's what you get when you allow free form classifications.

I suppose you could go through and automatically pull out relevent words but there is an art to indexing; “Dave Weiner” should not only be treated as one word for indexing, but it should instead be “Dave Winer” (typos and mispellings are another problem).

Sean, it sounds like you have a metadata habit that you may want to work on controlling ;) Perhaps if your software made the classifications more useful, so that they were always in your face, you would clean them up more often, or at least exercise more restraint in creating them in the first place.

With regard to automatically pulling out words, this is usually called "entity extraction". It's useful for things like spelling correction, but usually not reliable enough to use for metadata tagging.

You may have a point about my metadata habit (check out the source to one of my blog pages). But I do find the classifications useful as it's less data to slug through when searching for something than going through all the entries.

Besides, now that I have software that will actually list the classifications and entries that fall under each classification, it's amusing to see what some of them are (“Freedonian Feminists and Knitting Society”).

I suggest you look at TWiki - http://twiki.org

Its supports not only distinct namespaces but one extention even allows for heirarchies within those name spaces.

Various ways of categorizing are available, the most popular being the use of forms in the metadata. The forms can be richer as needed and wny namespace - "web" in Twiki terminology - can have a library of forms.

The deveopment web, for example, has evolved a policy that each topic has a single clasification: http://twiki.org/cgi-bin/view/Codev/TopicClassification
see the form at http://twiki.org/cgi-bin/view/Codev/TopicClassification#TopicEnd

If course the design also means that the labels on the forms are themselves hyperlink....

Forms are easily designed and specified:
http://twiki.org/cgi-bin/view/TWiki/TWikiForms
for example ...
http://twiki.org/cgi-bin/view/Codev/WebForm

Of course this is a design decision. I've built forms that allow of multiple categories, that the category label wikiword goes to a metasearch ...
metadata searches are part of TWiki as well.

TWiki has a lot of good ideas that should be looked into.

"First, wikis don't suggest refinements (other categories which could be intersected with the current category in order to narrow the search). [[ (yes they do) see http://twiki.org/cgi-bin/view/Plugins/SuggestLinksPlugin ] Second, wikis don't do hierarchical categories. [ (yes they do) see http://twiki.org/cgi-bin/view/Plugins/TreePlugin ] The combination of both of these features enables a qualitatively different kind of site."

I agree that the combination of both of these features enables a qualitatively different kind of site, which is why i use twiki from twiki.org. It has these features and hundreds more to choose from via add-ons and plugins.