In view of Jon Udell's post on Collaborative filtering with del.icio.us and continuing my last post
here, I wonder what it would take to create a tag-affinity feed:
something that will a list of my tags, see what items are associated,
for those items see how other users have tagged them and create sets of
related tags. (I feel the words "statistical clustering" rapidly
approaching this post).
For example, I usually tag RubyOnRails related stuff with a "rails" tag because, like duh it's Ruby
related. But how many other del.icio.us users have tagged it with a
"rubyonrails" tag or some variant? How will I find out all their links, besides
occasionally scanning manually? I could of course tag it as both "ruby"
and "rubyonrails", but that (a) annoys me, (b) doesn't cover alot of
cases, and (c) is not nearly geeky enough.
There's plenty of ruby libraries that can handle a lot of the heavy
lifting and one just needs to write the glue. I don't have enough time
for this at the moment*, so I'll just wish it on the web and wait a
couple of weeks until I find out someone had already implemented it :-)
* Pimki update
The reason I don't have time for the above at the moment, is that
Pimki2 is rapidly approaching alpha stage (with an ever decreasing
epsilon of Real Soon Now ⢠:-). I have most of the old functionality
ported over to the new Rails-based structure. There are also updates to
existing features and some new ones that I want to have reasonably
working before releasing as alpha. I'll post soon with more details.
This is the second stage of "make it work, make it right, make it fast"
development, so I am investing a lot more thougth and effort into unit
testing and integration, not just hacking stuff together... so it also
takes a bit longer.
Thanks for the 3,500 people who have downloaded so far, and the new
userstrying it out for the first time. Your input is invaluable!
Previous Topic
