Search relevance
Ethan Stock talks on his blog about the four fundamental ways to determine relevance in search results. These are:
1) Market/collective/algorithmic - Call it the law of large numbers, the wisdom of crowds, or statistical truth, but PageRank and similar "hubs and authorities" algorithms figure out, on average, what everybody thinks of a given site.
2) Expert/editorial - Roger Ebert on movies. Julia Child on cooking. Henry Kissinger on diplomacy. They've spent a lifetime figuring out what is the best, and the'll tell you so. Yahoo's roots are in this sort of directory/taxonomy/recommendation based organization.
3) Personal - Based on the user's past history and demonstrated interests, give them results that most closely match their tendencies. Amazon's A9 is the best current example.
4) Social - Based not only on the user's past history and demonstrated interests, but on those of all the people they know. This is a fascinating middle ground (possible sweet spot?) in between personal and collective/algorithmic -- you've got enough critical mass of opinions on key topics to hopefully eliminate outliers, but you've got a close enough connection to the user that you can bend your results to match their point of view.
The one that he left out that is getting important these days is location. All the major search engines have the concept of local search, which can retrieve results given a certain location either explicitly (user specifies which location) or implicitly (GPS device, Mobile network location etc). So for instance if I search for "weather", then the most important search hit should be the weather of the city I am currently in.
Medio is a mobile search company trying to incorporate location and personal data to provide answers to queries rather than just search results. The company came out of stealth mode earlier this year and has put up some information about their product on their website.
Links
posted by Gautam at
11/07/2005 04:33:00 AM
0 Comments:
Post a Comment
<< Home