twopointouch

One of the less controversial ‘good things’ about Web 2.0 is the tag cloud. There’s one on the top-right of this page. In years gone by, that would have been a list of the categories covered on this site. So why is the cloud better than a list? Two things occur to me:

(a) The compactness means it has an immediate visual impact. You can get a good sense in one glance of what this blog is about and whether or not you might want to follow it. While that cloud happens to be in alphabetical order, arguably size order would do an even better job.

(b) It’s a tool as well as a picture. Each of those words leads to a series of articles. Maybe you came here looking for articles about business. You can see I sometimes write on that subject at a glance and with a single click you can bring up all of them. (That is the reason I choose alphabetical order, by the way: so readers can look up topics more easily).

They’ve become ‘part of the furniture’ of Web 2.0 sites, perhaps, appreciated but not really considered. So I was interested to see the topic re-emerge this week on the O’Reilly Radar blog. Andrew Odewahn, director of the O’Reilly network had applied a tag cloud filter to Tim O’Reilly’s ‘What is Web 2.0‘ essay. He ended up with this:

And you can see one definition in a moment’s glance – a pretty reasonable one at that – of Web 2.0. It’s web software applications and services that deal with the links between data and users. You don’t even have to read the essay to get a good idea of the O’Reilly’s point of view of what is important in this sphere.

On the other hand, though, there are some big holes in this picture. It doesn’t mention anything about ‘collective intelligence’ or the ‘wisdom of crowds’, both of which are pretty central to the web 2.0 idea and the essay. Why? Because they are phrases and computers aren’t very good at dealing with language beyond single words. To a computer ‘collective intelligence’ is just two words; the ‘wisdom of crowds’ is three. It’s picked up the word ‘collective’ but it’s in small type. The algorithms used to generate clouds might be tweaked or primed by an intelligent operator to provide a ’seed list’ of terms, but making the creation of clouds automatic – so it can recognise key phrases without help – remains a bit of a holy grail.

So going forward, a few interesting alternatives have emerged through the comments. Mark Woodman is working on a ‘tag constellation‘ that analyses the top one hundred blogs to come up with a cloud representing the current ‘buzz’ across the blogs. This was about a story concerning Fujitsu making a very big hard drive. It’s also very interactive so clicking on the orbiting words would lead to other, connected stories.

Moritz Stefaner is working on another alternative. The colour intensity complements the words to show the extent of the buzz around particular topics. The cloud retains intelligence about what words relate to which topics, but fades them in and out according to whether or not they are hot right now:

Ted Shelton at Personal Bee is working on phrase analysis. Their site is a news aggregator and so identifying hot news stories depends on the recognition of phrases. ‘Big’ shouldn’t be recognised, but add ‘Brother’ and it’s probably part of a story. Add ‘racism row’ and you definitely know it is. Here’s a part of their view of the buzz around the topic of Web 2.0 today:

Of these three alternatives, it’s impossible to say that one is on the right track or that others are wrong. That will depend on the specific application they’re being used for, and the tastes of users. Different demographics will probably like one model more than another. Two things though, I think are absolutely vital for the next generation of any of these news tracker applications:

• Phrase recognition – This can probably happen automatically through language analysis for established stories, but may need to be done by hand for now for breaking news.
• More dimensions – we already have size, but currency and authority could really improve the ability to navigate such clouds.

Don’t know if you have seen the Quintura search application? It creates a mind map around search terms, letting users delve in and refine their topics on the fly. Sadly, it can’t do phrases and its language analysis doesn’t always work well. However, a new, improved Quintura for today’s blog posts from the people I read and the people that they’ve linked to is what I really want.

Technorati: folksonomy, tags, Web 2.0, website