Luke’s recent post challenging the design of search results pages reminded me that my original genre talk had another way to use document genres to improve the web experience.
When you use a search engine, say, Google, you get a list of ordered results. So, here’s what a search for “information architecture” returns.
Which is all well and good, if you want that particular tutorial, or Jesse’s particular set of resources. But, considering the 684,000 results, one gets the feeling that there could be more going on here.
Clusty, the hideously named search engine, uses algorithmic juju to identify related topics, and presents those with your results.
All right. I can’t argue that this is any better than Google’s results, because it’s unclear just how these topics are related. If you search on “information architecture” and then click “usability” or “web design,” just what are you going to get?
For my talk, I mocked up a Clusty interface that utilized genre as a filter:
The advantages, I hope, are clear. Utilizing genre allows the results to speak to whatever task the user brought to the search. If someone types “information architecture” into Google, they could be engaged in any number of tasks — self-teaching, looking for consulting help, looking for a definition, considering a career change, looking for a professional association, conducting research, etc. etc. Who knows? By offering genres as a filter, people can use that to narrow the results to those which are personally relevant at this time.
Now that would be useful!
PS. I hate typekey – by the time I get through the authentication, I have forgotten what I was going to post 😉 As you can tell from my very insightful comment.
your solution is lovely, but technically impossible at this time. The metadata isn’t there for this on the vast bulk of websearches. In a site-search, perhaps. Websearch is hard, as many panelists pointed out. really hard.
I’m there. It’s VERY useful to filter by genre. How do I get there though? Allow/force the search engine to figure it out? (I guess that’s the only option on the public search side.) Or is this another tagging exercise? Another kind of controlled vocab?
I’m on the private/enterprise search side of things and this kind of filtering would be great. Any ideas on approach or implementation?
Thanks for making me think though. 🙂
This idea is tempting but, I suspect, wrong.
It takes me back to the bad old days as an analyst, and as a buyer of analyst services. Analyst deliverables were typically posted online, organized by genre: executive summaries, newsletters, briefing papers, reports. What’s worse, the marketing folks typically made up catchy and meaningless names for these products, like Hot Bytes, so you couldn’t figure out which was which. If you’re looking for content on say, IM market share, you need to slog through every entry, particularly if the search engine was bad (and many were at the time). I’m not convinced that forcing the user to understand the vendor’s arbitrary packaging does much that’s useful.