blogrolls, tags, and similarity

Well, I couldn't upgrade the dKosopedia today, but I've had an idea in the back of my mind for a while and decided to look at it. This is kinda sorta related to tags, of course. Right now diaries get a single set of tags, which is suboptimal. Eventually, I'm hoping before dK4, but certainly as part of dK4, each user will be able to assign their own sets of tags to diaries; in other words diaries will have multiple sets of tags. (This is nothing startling, it's moving from the Flickr model to the del.icio.us model.) In addition, there's nothing that says that only diaries can receive tags; you should be able to tag other resources as well. If you think about it, we do that already with comments. Each comment can be tagged "Recommended" or "Troll", with visible counts of the numbers of users who have so tagged.

More...

Well, what else can we tag? How about users? Wouldn't you want to tag different users based on their interests or reliabilty or humor? I would. And then why not assign tags to yourself to reflect your interests.

Well, we sort of do that now, via the new personal Blogrolls. In a way, the blogs we put on our Blogrolls are proxy tags for our interests. We're tagging ourselves. Well, then, so who are we?

I took everybody's Blogroll entries as of 3 or 4 days ago and counted our entries. 1271 of us have blogrolls, with 10094 entries covering 3853 blogs.

Here's the top 25:

24	firedoglake.com	213
45	talkingpointsmemo.com	206
99	crooksandliars.com	199
29	digbysblog.blogspot.com	174
1	atrios.blogspot.com	156
56	mydd.com	135
160	boomantribune.com	125
159	huffingtonpost.com	118
58	salon.com/opinion/greenwald	106
102	juancole.com	104
161	myleftwing.com/frontpage.do	98
95	americablog.blogspot.com	93
32	myleftwing.com	87
253	thinkprogress.org	76
61	patriotboy.blogspot.com	75
290	rawstory.com	70
189	eurotrib.com	67
191	mediamatters.org	66
226	riverbendblog.blogspot.com	59
63	rudepundit.blogspot.com	57
28	pandagon.net	56
105	dneiwert.blogspot.com	56
78	talkleft.com	50
91	bonddad.blogspot.com	49
101	glenngreenwald.blogspot.com	44

The first number there is my internal blog id. Ignore it. You'll notice some blogs, like Glenn Greenwald's, have more than one url spec, which screws things up. I took off www. prefixes, and trailing slashes, but still...

Anyway, this is sort of a teaser, I guess. I'll put the whole list up on meta.dkosopedia.com somewhere, with a search form -- stick in a blog name and it'll return all those who have that blog in their blogroll. It would help if there was a more standard means of specification. How about myleftwing.com instead of myleftwing.com/frontpage.do, for example.

Then the next step is implementing a similarity algorithm. You put in a username and get back all the most similar users, based on blogrolls. All you interested in healthcare issues can find each other.

This is a fun thing, not even really useful as a proof of concept, but I'm hoping some of the code can be used later when determining similarities based on tags.

For the moment, a quick and dirty version will be here a little later tonight. It's dinner time.

•  damn

  the table got screwed up... I'll fix it later.

  by The Centerfielder on Thu Mar 22, 2007 at 04:51:29 PM PDT

  •  ok now

    by The Centerfielder on Thu Mar 22, 2007 at 06:12:45 PM PDT

•  That table is quite interesting

  Thanks for counting them. (manually? yeesh! tedious!) Do you know nos. 24 and 25 if "myleftwing.com" and "myleftwing.com/frontpage.do," and Greenwald's posts are mashed together?

  P.S. If you want this diary to draw more readers, you **might** want to consider altering the title. At least I, and probably others, see "blogroll" in a title and automatically identify it with "meta shitstorm."

  by righteousbabe on Thu Mar 22, 2007 at 04:56:24 PM PDT

  •  I have trained the mice

    in my computer to count, so they do the hard work.

    by The Centerfielder on Thu Mar 22, 2007 at 06:13:53 PM PDT

•  One question about blogrolls:

  there are other blogs?

  by Crashing Vor on Thu Mar 22, 2007 at 04:58:11 PM PDT

•  I don't know

  We have enough meta trouble here without being able to tag each other's IDs and individual comments.  I foresee trouble.

  ps.  very interesting.  great post.  I don't mean to negative.

  pps.  I read the first five a lot, but didn't put them on the blogroll since it seemed silly.

  by John DE on Thu Mar 22, 2007 at 05:03:06 PM PDT

  •  tag-related meta trouble

    Hopefully a lot of the "meta trouble" would disappear if tags were individual instead of global.

    Depending on how it was implemented, the tags you create would probably only visible to yourself -- which would eliminate troll tagging and the like.

    by Abou Ben Adhem on Thu Mar 22, 2007 at 05:55:51 PM PDT

    •  Yes, tags will be individual,

      and they'll be included in aggregates. But whether you want to make individual tagging public or private will be an option.

      by The Centerfielder on Thu Mar 22, 2007 at 06:12:20 PM PDT

•  Glenn Greenwald no longer posts @ blogspot

  he's only @ Salon nowadays.

  by Trix on Thu Mar 22, 2007 at 05:21:54 PM PDT

•  I've always hated the departmental org chart.

  All those solid and dotted lines that presumably lead you where you're supposed to go.  Somehow they make the world smaller.  I want to leave cube land at the office.

  This will be an appealing idea for many.  Occasionally, for me.

  by sele on Thu Mar 22, 2007 at 05:22:41 PM PDT

•  myleftwing.com/frontpage.do

  is the same as myleftwing.com

  by ek hornbeck on Thu Mar 22, 2007 at 06:29:05 PM PDT

  •  Ok you talk about that.

    "It would help if there was a more standard means of specification."

    Eh, I'm not so much into regimenting people.  That is what machines are for.

    by ek hornbeck on Thu Mar 22, 2007 at 06:35:29 PM PDT

    •  So

      you mean machines are for regimenting people? ;-)

      Yes, and that's the plan for tags -- maps from variants to a single tag. But it's harder work than I wanted to expend for an afternoon blogroll project. In this case it's just easier to rag on people for not taking data nerds into account.

      by The Centerfielder on Thu Mar 22, 2007 at 06:39:43 PM PDT

  •  Yes

    and calitics.com is the same as calitics.com/frontpage.do; and salon.com/opinion/greenwald is the same as glenngreenwald.blogspot.com. Tags have the same problem. Data is messy.

    by The Centerfielder on Thu Mar 22, 2007 at 06:35:59 PM PDT

•  Unrelated dKosopedia query

  The tags page and related pages (e.g., new tags) were not updated this morning. Any chance of it happening, or will we not be seeing any updates until after the move?

  by sardonyx on Fri Mar 23, 2007 at 07:36:03 AM PDT

  •  Ah, I wondered

    had an hour this morning for tag cleaning and always like to start fresh.

    by CalifSherry on Fri Mar 23, 2007 at 08:08:39 AM PDT

    •  I've been working

      on some of the extant tag jobs, and doing singleton cleanups using the excellent tag search page. But swatting the new daily singletons as they occur is vastly preferable to letting them sneak onto the list.

      I added a couple more jobs to the cleanup page: a guy we love to hate and a guy we love to like. How's that for balance? :)

      by sardonyx on Fri Mar 23, 2007 at 09:01:27 AM PDT

      •  I can get into that. Tonight or tomorrow morning.

        Can you answer my question about tag groups? Jotter's report includes a reference to Number of tag size classes. Where do I read about and see them?

        That tag search page is nifty.

        by CalifSherry on Fri Mar 23, 2007 at 09:54:46 AM PDT

        •  I haven't a clue

          Fortunately, I see you've already asked about it in jotter's High Impact Diaries diary today. If that doesn't get an answer, reply directly to one of jotter's comments for more visibility. :)

          by sardonyx on Fri Mar 23, 2007 at 12:35:30 PM PDT

      •  Gonzales tags are all good...

        for the moment. First name added, last name spelled correctly, other oddities all better now.   /happy dance.

        by CalifSherry on Fri Mar 23, 2007 at 07:49:26 PM PDT

•  Tag cloud question

  Jotter reports on the

  Number of tag size classes: 534

  I've not found reference to tag size classes in FAQ or dkosopedia. What are they, how calculated and can I see them? Pleeeese.

  by CalifSherry on Fri Mar 23, 2007 at 08:20:15 AM PDT

  •  Hmmm

    I don't know. I assume they're the number of counts. One "size class" is 1, another is 2, etc., with the largest size classes the numbers for the "George W. Bush" and "Iraq" tags.

    by The Centerfielder on Sun Mar 25, 2007 at 12:16:15 PM PDT