For example, our department that handles “garbage” or “trash” is called “Solid Waste Management.” I don’t know anyone that would think of that term if they were looking for a schedule of trash pickup times.

Has anyone out there tried this on a government site, and if so, what were the results?

See the wikipedia entry for more information and links: http://en.wikipedia.org/wiki/Faceted_classification

I realize that a tag set isn’t an index per se, but two key things you learn as an indexer is that you have to use terms consistently (e.g., pick Hawaii or Hawaiian, but don’t use both), you need to make sure you index broadly enough for your audience (e.g., Hawaii *and* palm tree), and you need each index term to have only so many entries under it or readers will be overwhelmed. This last one is more true in electronic indexes, such as in online Help systems, where you can assign index entries to pages to your heart’s content and not know until you look at the resulting index just how many hits each term has.

I find that tag sets, when users are good about tagging, are interesting and show a lot about how the tagger thinks. But they don’t yet replace good old search for me.

In a corporate context, I could see tagging working on an extranet where there are significant numbers of worker-contributed docs. (Each author would be asked to type a few words describing the document, and shared corporate jargon would probably help keep the folksonomy consistent despite smaller numbers.)

I’d also like to see tagging UIs evolve: There’s no reason a multi-level filtering UI couldn’t be constructed for something like Flickr, wherein you click on tags sequentially to narrow the result set.

Lastly, I’d be interested to see how effective a hierarchical folksonomy might be. Certainly there are parent-child relationships amongst tags in a system like Flickr.

It’s interesting that Josh uses the example ‘cameraphone’ to illustrate the usefulness of folksonomies. I would say that this example is better to illustrate the shortcomings of the technology. The tag ‘cameraphone’ is completely useless for users to retrieve information. Why would I care if the picture is taken with a ‘cameraphone’ or an ordinary digital camera? Or an analogue camera for that sake? This is a good example of focusing on the wrong thing – technology instead of content.

In Amazon’s case, the instructions for tagging are questionable. Sure, they tell you that if a book is about cheese, then tag it with ‘cheese’. They also tell you that if you’re considering it as a gift, you might want to tag it with the name of the potential recipient. But both tags are pointless (for everyone); if a book is about cheese, then it will have the word “cheese” in the title, or at least in the description. And unless I’m missing something, I cannot fathom a use for a collection of “items sold by Amazon that are potential gifts for people named Tom”? It might work for people to organise their own private gift lists, but Amazon makes no distinction, and for a public folksonomy it is absurd. By definition, Amazon is already highly catalogued and organised, and I question the need for any kind of user tagging on their site (beyond the “tech buzz” factor anyway).

The other example, Google video, calls their tags “labels”, and allows anyone to add them to any clip at will. On the surface, judging from the selection of tags on most popular clips, it seems like Google video users are adding inappropriate tags (‘xxx’, ‘adult’, ‘Zidane’ and so on) to every clip they see. Actually (and the last tag there is the clue) I think what is happening is that users are mistaking the “Add label” field for a search term box, and most of the tags are actually attempted searches. Of course it doesn’t matter _why_ people are adding the irrelevant tags, as they still significantly diminish the potential benefits of tagging.

So the results are unclear for Google video. Perhaps allowing uploaders to tag their uploads (à la Flickr) would be a better start than offering it up to random viewers.

Personally, I think that — at this stage in its life — tagging works best in a controlled environment, where the people entrusted with adding the tags are also the content owners/experts and where the concept of tagging is understood by all involved.