About a month ago, Tim Bray of Sun more or less stunned me by simply asking the question of whether or not tags matter? Last week he furthered my confusion by asking whether or not they work. From my increasing usage of tags, the answer was so obvious to me that I didn’t understand the question (it may well be that I still don’t).
I should admit upfront that it’s possible that I’m not the right person to look at the issue objectively, given that I have nearly 1300 tagged items in del.icio.us and over 400 tagged photos in Flickr, but neither do I consider myself a tag zealot. They’re useful in certain contexts, less useful in others. I’m certainly not running around trumpeting them as a replacement for Google.
But as much as I’d like to dismiss the question as the product of someone who doesn’t “get” tags, one obviously can’t overlook any commentary from Bray, let alone such a fundamental question. He’s not only sharp technically (yeah, this is the inventor-of-XML Tim Bray), he’s eminently practical in his approach (see his comments on the Loyal Opposition), and is not one to merely toe the party line (see his work on Coyote, comments on Solaris, etc).
But just from the way he frames his questions around tagging:
My question from last month still stands: Are tags useful? Are there any questions you want to ask, or jobs you want to do, where tags are part of the solution, and clearly work better than old-fashioned search?
I think I finally understand where Tim and I diverge with tagging. Tim seems to view tagging primarily as a competitor to traditional search, while I think the two are very different, though related, animals.
The critical ways search and tagging are different (in my usage):
1. Search is about information discovery, tagging is about information retrieval
2. Tagging (what is metadata but a tag?) is required for some search tasks (how – without tags – do I search and find a photo? Or an MP3 without ID3 tags?)
3. Tagging is by nature persistent, search is by nature transient (yes, I know search can be persisted, but it’s not typically)
4. Because of (3), tagging lends itself to collaboration (team tags, etc)
5. Because of (3), tagging lends itself to BI type reporting (see this, and imagine adding weighting, timelines, etc)
Ultimately, I think search and tagging are fundamentally intertwined, almost inescapably so. Tagging is declarative while search is implicit; it’s not a binary one or the other, I’m going leverage both to different ends.
But to make sure I answer his question, here are a few tasks I’d have trouble with had I not the ability to tag assets like music, bookmarks or photos:
1. Find all the pictures I took in game 1 of the American League Championshop Series (tagged answer here)
2. Find the articles I’ve already read in my research duties that reference Tim Bray and Sun (tagged answer here)
3. Find the band I’ve listened to most in the past week (tagged answer here)
I just as easily could have come up with examples that stumped tagging, of course, and that in the end is the point I’m trying to make. Search and tagging are different means to often similar ends, and provide us with an enhanced ability to retrieve material as well as manipulate it later for different ends.
For traditional information search type actions, desktop search may be a great fallback solution because we can’t and don’t tag everything that we might want in the future, but an explicitly applied tag does, IMO, improve the relevance of queries against that information store.
So I’m not sure if I answered the question any better than my colleague did, but either way I’ll keep chugging away with tags.