In search of navigation
Although there are only a handful of posts I’ve put up on thinkfetti so far, I have been trying to implement some navigation ideas I have. After several hours searching (and trying out alternatives), I’ve decided I will have to try again later.
TagClouds are interesting and visually appealing (and can be used for intrasite navigation), but I would like to be able to display them contextually; after clicking on one tag (“design”, for example), I would like the cloud to display its results for all the other tags present besides “design.” In other words, combining the flat orginizational structure of the tag cloud with the power of hierarchical searching. One article I’ve found describes a theory of tag clusters, which may be closer to the mark. And here is another thoughtful take on how tag clouds may evolve. The same author also refers to faceted classification systems, which may well describe the underlying structure I’m looking for.
Hierarchical categories by themselves don’t solve the problem, because with them, a target document is found by digging down through the hierarchy in the one way it is set up. For example, consider “education/design/music/Beethoven_improvisation”. The “Beethoven_improvisation” page could be found by opening “education”, then selecting “design” from the available options, then selecting “music” from the new list of sub-items, and then selecting “Beethoven_improvisation”. But you wouldn’t arrive at the page by selecting “music” then “design” then “education”, because the hierarchy wasn’t set up that way. And if “education”, “design”, and “music” were regular tags, you couldn’t use them to refine a search; clicking on any one of the tags would just open up a list of all posts with at least that one tag attached.
This touches on an important point about searching: a good search tool will help you find something you want, even when you don’t know in advance how it has been categorized or labeled. This is why physically browsing at a bookstore can be so rewarding, when constructing database queries may not be…
October 19th, 2006 at 3:19 pm
I like this tag cloud idea because it gives you, not only a list of what a text object is about, but how important each idea is to the collection. Imagine if library catalogs worked this way–you could see, for example, that your city library has an incredible biographic section but strangely little on gardening. If you have EndNote, you can do that with some companion software called RefViz. Run your references through RefViz, demote or promote the importance of words in which you’re interested, and see a 2.5D or 2D graph display of clusters of words and their frequencies. It’s a neat trick, and sometimes you find that there’s more going on in that text than you thought.
October 19th, 2006 at 4:47 pm
Thanks for the comment (the site’s first!), Bibliotechnocratissima (mixing a little Italian in with the Latin – hope you don’t mind)….
There are several initiatives underway to implement faceted searches in libraries, (such as http://sourceforge.net/projects/patacriticism and http://flamenco.berkeley.edu/index.html), but I haven’t yet seen faceted search combined with the appealing design of the tag cloud, which uses font size, color, etc. to add some dimensionality, as you’re saying.
I think a really well designed search interface will have to get past just displaying lists of words as links, without being so clever graphically as to be daunting.
And if it’s really, really well designed, it will easy to implement, too. So if the EndNote/RefViz thing – with formatted output – “just worked” (set itself up, had practically no learning curve for basic use), it could make a big impact on how non-experts found things (and experienced information), don’t you think?
October 24th, 2006 at 4:43 pm
Wow Doug. This made me long for my days of MARC records, Boolean operators and LC cataloguing systems. That’s saying something.
I really do like your idea of combining the tag cloud with a hierachical system. I’m sure you could get this going.
October 24th, 2006 at 10:53 pm
Sorry to incite a flashback, Rebcamuse…I don’t like to abuse friends.
Boolean would be much less painful if I just had to click a few times instead of remembering syntax (which of course differs on every database and search engine). It’s kind of you to suggest I could get it going, though – the amount of programming syntax I would have to master to set it up so I could browse with clicking instead of syntactically typing is a little daunting.
I would be pleased if some clever programmer would just implement it for all of us to use!