John Battelle's Searchblog

Naturally, the Newsweek article is not aimed at technical people. I can't say the other source is much easier to read though.

seriously

You might find this related article interesting:

http://digg.com/software/Why_current_tagging_sucks,_and_how_to_fix_it

Just skip over the machine learning algorithm at the end and it's quite short (in bullet form).

-- Dan

Wow, what insightful commentary....oh wait, I'm being sarcastic. Frankly, I have to say his opinion about tagging is not especially refreshing and is just all around cranky. While I am not sure that it is worth the hype, I do think that his problems with tagging are just plain silly.

What is the problem with synonyms and pluralization? Sure they make the results less pretty and a little uglier, but they make them more useful. It is safe to say that someone out will use the same synonym.

Tag spam could be a problem, but it isn't yet. That said, there are ways to deal with that. Just because tag spam is theorectically possible does that make the medium useless? Does link farming and blog spam invalidate google?

Also, is there a problem with blogs as a popular tag? The sites they point to are in fact blogs, so they are accurate. You could argue that this information is useless, but that point is just silly. If I click on blog, then I can see all of the blogs that have been tagged. By the fact that they were tagged indicates that someone at some point found that blog to be interesting. If I see that a ton of people found a blog to be interesting, then I might want to check it out. The blog tag just became useful.

Plus, if there was a user whose other tags were interesting to me, I very well might want to read a blog recommended by him/her.

Librarians need to stop being so cranky about tagging. It is really the best way to bring an overall browse experience to the internet. It might not be clean, pretty or heirarchtical, but it is pretty useful.

The debate between "hierarchy" and "tagging" should end now.

There is a functional isomorphism between "tagged semantic systems" (i.e. using features or dynamically generated microfeatures) and "tree-like hierarchies".

We should be exploring the two representations and building dynamic tree-hierarchy views that can also be viewed as taglists or tagclouds.

There is no contradiction - just two views of the same data - a bit like "wave vs particle" theory in physics

We should be building hybrid systems that can use both systems - and look at our data as hierarchy at certain times, but when we drill down we can use a tag-list or tagcloud view - and toggle between that view for any node in the "hierarchy".