Bookmark Management, Faceted Classification and Intelligent Agents | October 24, 2004
Tom Coates observations on bookmark management really do mirror my own experiences. All those years ago I started off just adding bookmarks to my favourites folder. The folder grew in size very quickly so I started adding sub folders with names like ‘Favourites’, ‘Web Design’ and ‘Fun’. As I added more and more links, I was forced to add categories ad hoc, and it wasn’t long before sub folders were needed.
Eventually I realised this add hoc method of categorisation just wasn’t working so decided to spend a whole weekend re-organising my bookmarks � basically the computer geeks version or reorganising your music collection. This proved to be a huge IA exercise but eventually things started to make a bit more sense.
One of the main problems with bookmarks is the classification model they use. Bookmarks are basically organised in a hierarchical structure similar to the classic computer file structure. This is great for small amounts of easily classifiable data, but quickly falls down with large amounts of multifaceted information.
For instance, take this site. Would you put it in a folder marked “blogs”, or possibly a folder marked “web design”? You may even put it in a folder marked “photography” if you were feeling particularly generous. Rather than use a hierarchical structure, things like bookmarks require a faceted classification. That way you can categorise a website by a variety of properties including topic, author, date published etc.
The other main problem is that of synchronisation. I quickly realised that my bookmarks would be useful at work as well as at home, so every now an again I’d export them from my home browser and import them into my work browser. Unfortunately because my home bookmarks form my master copy, every day I’m forced to email home half a dozen links to add to my bookmark list, which isn’t very efficient to say the least. I toyed with the idea of getting a .Mac account which allows you to sync your Safari bookmarks but decided it was a little overkill.
Instead I’ve been thinking about using del.icio.us. If you don’t know what del.icio.us is, it describes itself as a social bookmarks manager. It’s basically an online application that let’s you categorise your bookmarks with keywords and then store them online. Adding bookmarks is done using a favelet and if you don’t like the online GIU there is a simple OS X application called cocoa.licio.us available. If you want to sync your online links with your safari bookmarks you can do so with delicious2safari.
I’ve had a bit of a play with del.icio.us and it seems like a nice, simple idea. Whenever you bookmark a page you type in a list of keywords to describe it. However this means that each time you add a site, you need to think about it’s classification and what meta data you want to add. This really can cause a huge burden and forces people to create their own, ad hoc classifications and essentially their own controlled vocabulary. So while I like the concept behind del.icio.us, even if I did know how to upload my existing Safari bookmarks, I don’t have the time or inclination to go through 2403 links and classify them.
What’s needed is some semi automated classification process. Toms suggestion of using the sites Keyword meta data is an excellent one. Rather than force the user to classify each site they bookmark, you get the author to classify it. Another way to handle this would be to use pre-existing user data. When a user adds a bookmark that is already in the system, they would be given the option to add new keywords or choose from a list of existing keywords being used to describe the site. That way you’d end up organically creating a controlled vocabulary.
When personal computers debuted all those years ago, nobody really knew what they’d be used for. I doubt when the first hierarchical file systems were conceived, their creators had any idea that an average home computer would be able to store 60GB+ of data. Being lumbered with such an inflexible file system, software manufacturers came up with the idea of intelligent agents. Rather than hunt around in the file system to find information, you would use smart programs to sift through the data. An excellent example of a current intelligent agent would be iTunes.
Apple realised that people were likely to have thousands of music files on their system and wouldn’t want to organise them by album/song name alone. So they created iTunes, a basic intelligent agent that allows you to navigate your music library in a variety of ways. You can create simple playlists that you drag and drop items into. However you can also create smart playlists that help organise your music intelligently. For instance, most songs contain a certain amount of meta data like the album they came from, the artist and possibly a genre. iTunes can add it’s own meta data like when it was added to the collection, when it was last played and how many times it’s been played.
Imagine if you could do the same with your bookmarks. An intelligent bookmarks agent could analyse the keywords and even the content of the page to come up with a variety of classification facets. It would know that a specific page was from a specific site and who the author of that page was. So if you wanted to you could create a list of all the articles by Jefrey Zeldman you’ve bookmarked or all the articles on ALA you like. If this intelligent agent was linked into your browser you could create a “Favorites” bookmark list based on how many times you’ve visited that site in the last 6 months. Link it in with an RSS feed and you could create a list of all the sites updated in the last week. Conversely you could clean out the garbage by creating lists of sites you’ve not visited in over a year or have only visited once. You could even get your smart agent to ping these sites to see if the pages even still exits.
Posted at October 24, 2004 2:40 PM