nacturation's Journal: Slashdot tag censoring

The recent story on female shark reproduction really drove home what I've been seeing in others comments recently. Slashdot is now quite obviously eliminating certain tags, such as:

http://slashdot.org/tags/lasers (only one shark article?)
http://slashdot.org/tags/frickinlasers (zero found?)
http://slashdot.org/tags/fud (zero found!)
http://slashdot.org/tags/notfud (zero found!)
http://slashdot.org/tags/slashvertisement (zero found!)

Not that I don't blame the effort to increase the quality of tags, but it seems like all the historically popular tags have been retroactively wiped out rather than taking the approach of having a higher threshold for future articles. Tags such as yes/no/maybe are of little value other than a popularity war at the time the article is active and it seems logical to exclude these as simply tag trolling. However, what if someone wants to do a search on what the Slashdot community thinks are slashvertisements, or are articles spouting FUD? Just can't do it, and a potentially useful zeitgeist measure is lost.

So once again, Slashdot jumps the shark? :)
 

AH yeah

[grub]

Several weeks ago I posted (maybe as an AC?) in an article about one of the *AAs and wrote "Tag this 'thievingcunts'" Sure enough it hit and there it was on the main page. Doesn't show up in a search now though.

Re:

[RailGunner]

Neither does the Evolution article I got tagged "ButtSexWithFishSquirrels" from the South Park episode on it...

Re:

[nacturation]

Several weeks ago I posted (maybe as an AC?) in an article about one of the *AAs and wrote "Tag this 'thievingcunts'" Sure enough it hit and there it was on the main page. Doesn't show up in a search now though.

And surely that's one for the record books! I remember that one... and I think you posted that one non-anonymously -- would that be "nonymously"?
 

Re:

[grub]

Found it! [slashdot.org] I did post as AC. Interesting that the tag is there in the article page but when you click the link the search returns nothing.

Some comments

[jamie]

Hi, I'm the Slashdot programmer working on tags and scoring. I've actually been working on some code related to this all last week and I hope it'll go live next week.

The big picture is that we're working to find ways to assign users appropriate values to signify how we rate their contributions to the tagging system, which includes the firehose up/down voting and soon moderation as well. The algorithms are a little complex.

In particular, the current code doesn't distinguish between "how helpful a user is in the firehose" (our chief focus at the moment) and "how helpful a user is for tagging stories" (which is what you're talking about). That should come next week. Meanwhile, what brings tags onto the homepage is a different metric, one whose raw numbers are on a different scale than the other numbers, so not a lot of tags are making it to the homepage lately. That's basically just a bug.

To make things more complicated, the algorithm that tags.pl uses (at /tags/foo) to determine which objects to pull up is different from the algorithm used to actually put "foo" (or not) on a story for article.pl or index.pl. I will need to revisit that too and make those jibe, but I haven't even looked at tags.pl yet. So you can blame me for that part too. It's gonna get even more complicated once our search engine gets revised.

We haven't retroactively wiped anything out. The data's still there, and most of it is really useful and we intend to float the useful tags to the top-5 list -- that should happen retroactively in fact.

As you may have noticed, we manually remove some tags from being rendered onto stories for display by index.pl and article.pl. If users apply unrelated tags to stories, and we notice, we'll take them off. Those top-5 tags that get displayed for a story are intended to be used for categorization, loosely interpreted (by me anyway) to mean description, valuation, and related commentary, with the main criterion being "would anyone ever search on it." For example, "mafiaa" is both commentary and description of what a story is about, and we certainly allow that (right now I'm only seeing 2 stories tagged that way, but there are surely more; see previous paragraph). And (once the problems are fixed and display is improved), /tags/mafiaa could indeed be a useful listing of stories which our users have tagged as relating to the RIAA and MPAA and which they have made a particular value judgment about. It's even remotely possible (no pun intended) that someone might do a search at /tags/lasers to pull up the remote-controlled sharks story (that's kind of the boundary of plausibility). Even /tags/yes and /tags/no make for an interesting data set, so they stay.

However, nobody's ever going to search on "thievingcunts" to find stories. One-off commentary by definition isn't useful for that purpose. The meme-joke-of-the-hour is only funny for about that long and after that it's just -- from our point of view -- clutter in the folksonomy.

So while you're perfectly welcome to tag stories however you like for whatever purpose you like, and you'll always be able to browse your own tags as bookmarks at /~yourname/tags, we're not going to use one-off jokes in the places where we're trying to cultivate categorization.

I don't see this as censorship. Users are welcome to express themselves in comments on a story, at great length in fact. If you have something to say that isn't right for a comment, you can post journals on Slashdot too. As far as tags go, you're welcome to tag any stories with any tag you like, and you can view those at /~yourname/tags. Where we find your tags useful we'll include them on the homepage and for searching, and where not, we won't (time, attention, and my ability to code algorithms permitting).

Re:Some comments

[nacturation]

The "how useful is it for searching" criteria can be determined objectively by how many actually search for it. Maybe any tag gets an equal chance to show up but they have a fixed half-life assigned to them. The search frequency can directly influence the length of the half-life. So if "haha" and "itsatrap" and others make it into the tags list on the main page, that lets the community have their inside joke-du-jour. But over the long-haul, the half-life ensures that the tag will eventually die out unless it's balanced by searches for other stories with that tag.

For this to be effective, I think a "search by tag" feature should be promoted on the main page. Clicking on an existing tag to find related stories should be given limited weight, but independent searches for those tags could have a much higher weight assigned to them. Also provide RSS feeds for tags. Those who want to track "slashvertisement" articles can subscribe to the feed and publish summaries on their blog or whatever. You could even individually encode links back to the articles so that the more clickthroughs an article from the tag's RSS feed gets, the more weight is given to the tag.

Such a strategy should further the interests of the community -- let them have their laugh, but over time the jokes fade into obscruty. It also furthers the interests of Slashdot -- the useful tags will always bubble to the surface and become more relevant over time. If done right, the entire system would be completely automated and would only require tweaking of the algorithm to fine tune it.
 

Re:

[jamie]

The "how useful is it for searching" criteria can be determined objectively by how many actually search for it.

You know, that's actually a damn good idea. We're not tracking any correlation right now of search terms with existing tagnames, but we should.

Half-life sounds like a decent algorithm too, at least at first blush. I'll have to think that over. But we should definitely start logging which tagnames actually get searched on, so we can decide how to use that information later. Thanks!

For this to be effective, I think a "search by tag" feature should be promoted on the main page.

It'll be integrated with the main search (it's all being rewritten).

Also provide RSS feeds for tags.

Yep, that's on the to-do list for sure.

Re:

[nacturation]

Yep, that's [RSS feeds] on the to-do list for sure.

Awesome!

Half-life sounds like a decent algorithm too, at least at first blush. I'll have to think that over. But we should definitely start logging which tagnames actually get searched on, so we can decide how to use that information later. Thanks!

I'd be interested to see something like a half-life algorithm in use as it seems like the natural way to do it. If you decide to use it, be sure to reserve some /. swag for me. :)

Re:

[bjelkeman]

So if "haha" and "itsatrap" and others make it into the tags list on the main page, that lets the community have their inside joke-du-jour.

I actually think this is very important. The thread that went "off thread" [slashdot.org], is a good explanation of what I think about this too. The tags without the "short hand review tags" such as "dupe", "duh" etc. are essentially useless for the quick reading of the story. It substantially increased the speed of which one could consume Slashdot and understand what the content is about.

Re:

[metamatic]

You seem to be assuming that the only purpose of tags is searching.

It used to be the case that if I saw an article was tagged "troll" or "bullshit", I would know that the majority had recognized that the article was the latest Dvorak article or a bunch of made-up pseudo-facts, and I could skip it entirely. Now I have to wade into the discussion.

So basically, by censoring tags you stopped them from being useful to me.