Comments on: Tagging and long tails

By: EconoMeta » Blog Archive » The long tail tagging the dog

EconoMeta » Blog Archive » The long tail tagging the dog — Sat, 04 Jun 2005 01:52:29 +0000

[...] onalization and personal data The long tail tagging the dog In a previous post, I mentioned some interesting graphs that could be made from public URL tagging data such as th [...]

By: Adam

Adam — Wed, 18 May 2005 21:47:01 +0000

Thanks for the comment. It looks like you created a new chart of number of tags by number of users who use that many tags. This wasn’t one of the potential charts I mentioned; this distribution would show whether the long tail of users assigning a high number of tags cumulatively comprises more users than those assigning a smaller number closer to the average. From this sample, at least, it looks like this isn’t the case; most users assign a number of tags close to a median of somewhere around 50.

I’d still love to see URLs by number of times a tag was assigned, or tags for a specific URL by number of times a tag was assigned.

By: Phil

Phil — Wed, 18 May 2005 11:22:05 +0000

Thanks for flagging this up. Clay’s “long tail” graphic looked so ‘right’ that it didn’t occur to me to question whether it said what it appeared to say.

It doesn’t, frankly. If you re-crunch the numbers Clay was working with, you do end up with a huge left skew and a long-ish (although bumpy) right tail – working from ’0-25′ up to ’550-575′, the figures are
15,15,7,7,3,4,2,3,0,1,0,0,1,1,0,2,3,0,0,0,0,1,0

But the skew is towards the *low* end of the distribution – what Clay calls the ‘long tail’ – and (perhaps more importantly) that’s a tiny, tiny sample.