Sep 24, 2001: More on Smarter Search Results
Greetings from deep in the heart of Texas, where I'm giving a talk tonight at UT and enjoying the fine facilities and hospitality of the LIS program's computer lab.
In my last entry I discussed the nice job that HP's information architects did with integrating manually-derived results with automated search results. A couple of you wrote in to comment on this approach.
Matt "blackbelt" Jones was "excited and encouraged... for reasons that will become clear in a couple of months". Reasons which he did not not expound upon. But which I will be vigorously nagging him for until he coughs them up. Don't play with us Matt...
Matt also pointed me to a CHI presentation by Microsoft's Susan Dumais, Edward Cutrell and UC Berkeley's Hao Chen: "Optimizing Search by Showing Results In Context". In a nutshell: "Our user studies show that all Category interfaces were more effective than List interfaces even when lists were augmented with category names for each result". So categorized results are a Good Idea, even when categories are automatically generated. Categorized results provide users with greater context for the focused results that search engines provide; this conclusion seems to support my intuition that there's lots of room for improvements in search results presentation.
On the other hand, you might also argue that the Northern Light search engine is not setting the world on fire. How many of you use those Little Blue Folders regularly? Personally, I generally find them useless. And I think it's because they're automatically generated, not manually. But that's just me.
Anyway, here's what I think is happening. Northern Light and other good search engines are losing out to Google. Northern Light: interesting approach to search results presentation. Google: interesting combination of a pattern-matching algorithm for relevance determination and a link analysis algorithm for popularity determination (and probably some algorithm that combines the two). Google: more successful because it gives users a "better" first pass retrieval, though they might also benefit from a Northern Light-style results presentation. So perhaps an evolutionary advance in the application of search algorithms trumps a more revolutionary advance in search results presentation?
Well, what about a hybrid? Well, there actually seems to be one that fits the bill perfectly: Teoma. Check it out and you decide if this is the future of information retrieval on the Web. I sure hope so, because I bought stock in Teoma's new parent, AskJeeves, early in 2000. Ouch. Anyway, it'll be interesting to see what comes of merging Teoma with AskJeeves' approach, another neat idea that hasn't set the world on fire.
Back to Bloug reader comments: Stanford's Lisa Chan used to work at AltaVista, and points out that AV did what Hewlett-Packard is doing now. It was called "AV Recommends" and is no longer available. Why? Because AV laid off all those smart but expensive humans who generated these smart recommendations. Lesson: you can't classify or otherwise manually improve access to all web content. Something that many of us have been grumping about for about ten years now. But HP shows us that you can manually add value in narrower, selective environments, like the corporate intranet. In fact, HP ought to contact Lisa to track down all the other AV alumni and put'em all back to work...
email this entry
Comments are now closed for this entry.
Comment spam has forced me to close comment functionality for older entries. However, if you have something vital to add concerning this entry (or its associated comments), please email your sage insights to me (lou [at] louisrosenfeld dot com). I'll make sure your comments are added to the conversation. Sorry for the inconvenience.
©2008 Louis Rosenfeld LLC. All rights reserved.