Abolish tags, say hi to relationships
March 18th, 2010We’d all love it if more news websites simply started tagging their content. But when you think about it, tagging is a very primitive way of saying how things relate to each other. It says “this article has something to do with this concept or thing”. They don’t tell you exactly these things relate. They also don’t tell you whether the the thing the article relates to is a topic, a person, an organization, an event or a location. That’s okay for humans because we can guess, but computers can’t, and we’re really going to need computers to aid us in bringing context &emdash; even to do trivial things like provide an index of all the people that a site has content on.
With just a little more effort on behalf of the writer, it’s easy to specify real relationships in familiar triplets: critiques , contains an interview with , revolves around .
This information could be used to make search queries more intelligent, and to make sure topic pages are more than just link dumps that leave the reader figuring out how exactly the linked content relates to the topic at hand.
The New York Times already splits up its tags into four categories:
- (Des) = subject descriptor/heading
- (Per) = person
- (Geo) = geographic location
- (Org) = organization
Let’s take that system one step further and instead of just splitting up tags into different categories of tags, treat persons not like tags but like persons. Tags are just labels, persons have names, birth dates, profile pictures, things they’ve said (quotes), personal websites and so on. Ditto for organizations, locations and events.
I know of no CMS that supports this out of the box, which is understandable because being able to say “this sorta relates to that” is sufficient for most intents and purposes. It just doesn’t cut it for news websites, though.
I’ve just posted an expanded version of this idea over at http://stdout.be/2010/tags-dont-cut-it/ for those that might be interested.
@Stijn: I’ve been following your posts here and your blog re relationships; excellent suggestions. We’re working on implementing your ideas on our local news site, The Hawaii Independent. I’ll let you know how it goes. FYI, we’re running on ExpressionEngine, which has a very flexible templating system and makes it easy to relate posts to each other. We’ve got one weblog for all of our “stories,” and then separate weblogs for an index of people, and index of organizations, and index of events, and an index of topics and themes. We’re using a wonderful relationship plug-in called Playa to interrelate our story posts with respective people, orgs, topics, etc. It’s going well — the next challenge is to train our staff and readers to understand and take advantage of this architecture. I’m grateful for the time you’ve put into thinking this through, and we’re happy to (try to) implement it.
Ikaika Hussey
Publisher, The Hawaii Independendent
I agree heartily that tags are a very primitive way to organize information; but even more contextual and categorized tags will likely be limited. They’ll help search, but what we really ought to be doing is looking at the structure of the information we create, so that we optimize the form for the content.
For example, imagine a comprehensive stock market report or sports story that’s micro-tagged to within an inch of its life; it’s still not as useful as a table of stocks or a box score. That’s not to say the stories are useful; only that the tables, by virtue of layout and structure, also have value.
A lot of the discussion about search, and information, is centered around text (and to some extent images) in the narrative form and how to make it more contextual or searchable, etc; that’s important, but so too is thinking about the form of that text, and whether it should be captured more in tabular or other data structures, ala Politifact.
I muse more on this at my blog, structureofnews.wordpress.com