Tag-tree: What and Why:

* Bountyup.com (see my crazy startup blog thread, soon to be written) uses the now-ever-popular tagging metaphor for classification of items. However, I *still* want to be able to pretend to have a traditional taxonomy on the front page, represented as a tree (roughly speaking). Rather than build one and manually classify all bounties, or (even worse) force the *users* to classify them, I have a few choices:

  a) Map tags to categories in some one-to-many or combined-key-to-many sort of way
  b) Show every tag as a single flat category
  c) Make a tree out of the tags

I’ve decided to try and do the latter (yes, I realize this is one of those impossible web2.0-type dreams, since by their very nature folksonomies and taxonomies are different beast. I’m going to try anyway, and at the very least this will give me a starting point for (a) above).

My proposed algorithm:

1. Most commonly used tags become the top-level folders.
2. Tags used in conjunction become secondary folders.
3. Items tagged with a secondary tag and no primary tag will be disambiguated, if possible, by the following steps:
    a. If other items have been tagged by the same user with this secondary tag, as well as a primary tag, the matching primary tag will be assumed.
    b. If not, (eventually) content analysis will be used to disambiguate (full-text index of other secondary-tagged content to determine the appropriate primary tag).
    c. In the meantime, show the item in both locations, and possibly flag for manual review. (Maybe a cool tag-supporting UI that suggests possible primary tags to a secondary tag?)

Drawbacks, Advantages, and other Characteristics of the Approach:

* The hierarchy will constantly be shifting, since a secondary tag will become a primary tag simply by becoming more frequently used than the primary.
* Perhaps this should start out as an advanced or even admin-only view of the data?
* It’s possible (likely?) that there will end up with either a.) thousands of primary tags, or b.) thousands of ambiguated items with only secondary tags.

Let’s see how it goes.

Blogged with Flock