Richard has some very nifty ideas up his sleeve for the next iteration of his site. Some of these are design-related and some are technical. He just gave a peek into the technical side of things by explaining how he’s using tags to tie content together. Not just any old tags, mind: machine tags.
You may remember that Flickr rolled out machine tags a while back. That’s their name for what’s basically tripletags; tags that take the form of
namespace:predicate=value. There’s some tight integration between Upcoming and Flickr using the machine tag
upcoming:event=[ID]. You can see a looser coupling (one way rather than bi-directional) in the recently-updated events section of Last.fm which uses
lastfm:event=[ID]. As an example, take a look at the page for a Low Lows concert I went to and took pictures of.
Richard is making use of machine tagging to associate his Flickr pictures with his blog posts. He’s also planning to use Amazon’s API to associate ISBN numbers with blog posts, raising the question of which namespace to use:
We therefore need a triple-tag version of the ISBN tag, and here’s my suggestion:
iso:isbn=0713998393. ISBN is a standard recognised by the International Organisation for Standardization (ISO) so I thought it made a certain sense for ISO to be the namespace. Other standardised entities could be tagged in a similar way, such as iso:issn=15340295.
Seems like a sound idea to me. I might experiment with machine tagging reviews here in that way and then pulling in complementary information from Amazon.
But that’s for another day. For now, I’ve gone ahead and integrated Flickr machine tagging here… but this works from the opposite direction. Instead of tagging my blog posts with
flickr:photo=[ID], I’m pulling in any photos on Flickr tagged with
Now, I’ve already been integrating Flickr pictures with my blog posts using regular “human” tags, but this is a bit different. For a start, to see the associations using the regular tags, you need to click a link (then the Hijax-y goodness takes over and shows any of my tagged photos without a page refresh). Also, this searches specifically for any of my photos that share a tag with my blog post. If I were to run a search on everyone’s photos, the amount of false positives would get really high. That’s not a bug; it’s a feature of the gloriously emergent nature of human tagging.
For the machine tagging, I can be a bit more confident. If a picture is tagged with
adactio:post=1245, I can be pretty confident that it should be associated with http://adactio.com/journal/1245. If any matches are found, thumbnails of the photos are shown right after the blog post: no click required.
I’m not restricting the search to just my photos, either. Any photos tagged with
adactio:post=[ID] will show up on http://adactio.com/journal/[ID]. In a way, I’m enabling comments on all my posts. But instead of text comments, anyone now has the ability to add photos that they think are related to a blog post of mine. Remember, it doesn’t even need to be your Flickr picture that you’re machine tagging: you can also machine tag photos from your contacts or anyone else who is allowing their pictures to be tagged.
I realise that I’m opening myself up for a whole new kind of spam. But any kind of spam that requires namespaced tagging on a third-party site is pretty dedicated. If someone actually goes to that much effort to put a thumbnail of an inappropriate image at the end of one of my blog posts, I probably wrote something particularly inflammatory in that post—which would make the associated thumbnail a valid comment, I guess.
Here are some examples of posts I’ve been machine tagging on Flickr:
Once again, like Upcoming and Last.fm, these are event-based. But the machine tagging would work equally well for location-based posts. So when I go up to Scotland next week and blog about it, I (or you or anybody) can then go to Flickr, find some nice pictures of Edinburgh and using the adactio namespace, associate the pictures with the blog post.
It’s a strange mixture of RESTful URLs here and taggable objects there.
If nothing else, this will be an interesting experiment. Machine tags don’t have the low barrier to entry of regular tagging but they aren’t as complex as something like RDF. It might be that they hit the sweet spot between accuracy and ease of use.
Oh, and if you find any Flickr pictures related to this blog post, tag them with