Peter O'Kelly sent me a link to an interesting post on Read/Write Web, which pitches Search 2.0. It's a bit muddle-headed, as it talks about three generations of search technology and then promptly calls the third generation Search 2.0.
- First-generation search ranked sites based on page content - examples are early yahoo.com and Alta Vista.
- Second-generation relies on link analysis for ranking - so they take the structure of the Web into account. Examples are Google and Overture.
- Third-generation search technologies are designed to combine the scalability of existing internet search engines with new and improved relevancy models; they bring into the equation user preferences, collaboration, collective intelligence, a rich user experience, and many other specialized capabilities that make information more productive.
In fact, in a posted comment, Danny Sullivan takes the author to task for preferring marketing buzzspeak to clarity:
The third gen has commonly been considered the combination of personal data -- either refining results because of your own past history or that of others. Lump it all into social search, and that's your third gen. And since it's third gen, please call it Search 3.0 if you must and don't force it into a Web 2.0 world just to try and mesh some Web 2.0 companies into a Search 2.0 framework.
And, of course, he's exactly right.
Besides slamming misleading naming, Danny made another good point: unlike the other search engines named (Swicki, Rollyo, Wink, Lexxe), Vivisimo does not take the social aspect into account. Back in January 2005 I had a briefing with Raul Valdes-Perez, the CEO and co-founder of Vivisimo, and we had a five-minute argument about this. For the past several years I've felt that most search engines have neglected user profiles as a way to discern relevant content.
Think about it. Relevance is defined by the user. A search engine can categorize up the wazoo, but unless it knows that this naturalist is interested in apple the fruit, and not Apple the computer, it may suggest irrelevant content. Raul got quite adamant about not using user information to rank search results, citing privacy issues and adding a certain amount of handwaving. OK, whatever. For I view ignoring user profiles to be a fundamental design flaw in today's search market.
If you look at the three generations mentioned above, they follow a progression. Generation 1 looked at page content. Generation 2 looked at link analysis. This is essentially profiling content based on authority; the more sites linking to it, the better, more important it must be. Generation 3 finally addresses the Generation 2 imbalance. We may know a lot about the content, but unless we profile the users and distill what they're interested in, we'll have a hard time figuring out relevance.
So while the whiz bang interfaces mentioned in "Search 2.0" have a certain amount of value in that they personalize the interface for each user's way of working, let's not get that type of personalization mixed up with the deeper personalization available through user profiling. In terms of priority, it's more important to improve relevance than it is to improve the UI. Amazon.com gets it, and has for years. Microsoft gets it, with its forthcoming Knowledge Network. It would be nice if Vivisimo and others got it as well.
Comments