Log in

No account? Create an account

Previous Entry | Next Entry

If you use tools like Flickr, del.icio.us or Technorati, then you'll be familiar with the benefits of what have come to be known as folksonomies. All these services allow you to create your own tags for classifying content. User-created taxonomies work well in social networks, where there is an tendency to clarity driven by the ethos of the network. Flickr is a case in point, where specific image tags have become fashionable tools for sharing themed images without creating artificial groups.

It's quite simple to create your own folksonomy. Say you're using Flickr as a photohost, all you need to do is upload your pictures and then give them a series of tags that describe what you've uploaded. A picture of a man standing on a beach in the sun might end up tagged as "fred beach cambersands sunny". Four tags that really only mean something to you. However someone searching Flickr for pictures of sunny beaches would be able to find your picture, as their search would pick up the "sunny" and "beach" tags.

Folksonomies are a far cry from the complex taxonomies used by knowledge management systems, where there's considerable research into defining information hierarchies. But there's one thing about folksonomies: they just work.

That's fine. In fact that's just peachy!

The Ronseal "it does what says on the tin" approach is one I like. I'm as guilty of anyone as producing my own idiosyncratic tagging schemes - and I'll often have to go back to edit some consistency into my tagging.

However there's one area where I'd like to have some sort of constraint - and that's in what I've come to think of as "geosonomies": tags that relate to specific places. It's easy enough to find a tag on Flickr for London, but what if you're trying to track down an image of a specific place in a specific country, and you're not sure of the exact spelling. Surely it should be possible to start at "Belgium", and then use a map to find the place you want pictures from?

Implementing geosonomies should be relatively simple. All they do is imply a second level of metadata that says "place". They can be extended to manage well understood hierarchical structures - continent:country:province:city/town - so a tag "London" could be associated with additional hierarchical metadata: "Europe:UK:London". Geosonomy tagging can remain as free as the standard folksonomy approach - the additional metadata should be added automatically if the user identifies a tag as geographical. If the place name is unknown, the user can be prompted to add hierarchical information. We shouldn't expect services to build complete geographical databases when users can add places themselves.

Turning folksonomies into working metadata is an interesting problem for the next few years. Perhaps the geosonomy approach is worth considering as a place to start.

Nokia's Lifeblog is a step in this direction, and as more and more digicams and mobile phones start linking to GPS, geographical information is going to become part and parcel of our everyday online lives.

After all, we all like to know where we've been...

Thoughts sparked off by a walk in the park with marypcb


( 5 comments — Leave a comment )
Mar. 20th, 2005 12:00 pm (UTC)
it does seem to me that the online photo sharing sites are a good place to do this because of the benefits being so obvious that people would do the work. it still needs to be lightweight; a tickbox to say if a tag is geographical (I could use Seattle to refer to a style of music), an automatic search for matching placenames and the offer to enter a new location if necessary. But I think you need a basic database to start with because the general geographical opinions of people as expressed on the visited countries map discussion are so varied (read, erm, wrong!). My favourite comment is that 'countries need to be recognised by a global body like the UN and UEFA does not count'
Mar. 20th, 2005 03:37 pm (UTC)
I disagree. Folksonomies are generally based around nice fuzzy concepts that have no intrinsic heirarchical nature - taxonomies are usually based around some kind of concrete decision system (i.e. what is the phylum, kingdon, etc). Geographical systems seem to me to be to be mostly a case of the latter.

You pinpoint the location as closely as possible, and then it should be automatically recognised as being part of Edinburgh/Central Scotland/UK/Europe. Once I've pinpointed myself as being as a particular latitude/longtitude I should then be able to do searches without myself having to define the associations of that location (unless I want to have personal area definitions, obviously). GIS systems already do this very well (When I did an Arcview training course they showed exactly how easy it was to "Find all pubs within 500 feet of route X".
Mar. 20th, 2005 09:40 pm (UTC)
GIS is too accurate for what I'm proposing (though, yes, it is a logical endpoint). What I'm flailing towards is something that is fuzzy, yet hierarchical and doesn't conflict with the rest of a folksonomy...
Mar. 21st, 2005 12:22 am (UTC)
You don't _have_ to be accurate with a GIS. Once you've defined your areas then saying "London" will match onto an area and pick up the predefined relationships. Or at the least ask you "By London, do you mean this blob here in England, that blob there in Texas..."
Mar. 24th, 2005 09:00 am (UTC)
what's a fuzzy subset of a strict taxonomy? My original thought was mining the folksonomies for the overlap with taxonomy to get a stream of semi-structured info but the tagging is always going to trend to fuzzy rather than strict (to placename rather than postcode, say). What do we need to cope with this? Without absolute location you're getting a fuzziness anyway,because you can't do 'what's nearby' without having real locations that users aren't going to enter. If the EXIF data for the photo had GPS/lat-long sure - but there are bazillions of photos that don;t and won't. plus there's what's London vs what's Putney - my photo might be looking over the river *from* Putney *to* Fulham. Where does that go in the ultra-strict taxonomy?

GIS systems are still fairly proprietary and expensive (there is no UK equivalent of those public US maps), as are the Point of Interest databases retrieving those queries. Pulling out the folksonomy-tagged data into a taxonomy would give us more open source locationish data to make up for it.
( 5 comments — Leave a comment )