Wednesday, May 07, 2008

Finding the location of interests and objects from search logs

A paper at WWW 2008, "Spatial Variation in Search Engine Queries" (PDF), by Lars Backstrom, Jon Kleinberg, Ravi Kumar, and Jasmine Novak offered many clever examples of using where people are when they do a web search both to determine when interest in a topic is geographically isolated and to estimate the physical location of objects.

The most dramatic example the latter in the paper was estimating the location of a hurricane over time by the queries for the name of the hurricane. It worked surprisingly well.

I intended to write much more about this paper, but Danny Sullivan already covered much of what I planned to say in his excellent post, "Yahoo Paper: Finding The Local 'Center' Of Search Queries".

Danny excerpts many of the pictures from the article and comes to the same conclusion that I did, that it is fascinating, motivating paper, but it leaves open the question how to apply this to improve local search.

Please see also the papers at the LocWeb Workshop from WWW 2008, several of which also looked at using IP addresses in search logs to determine local intent.

1 comment:

Pranav Dandekar said...

One way I can think of using this data is to personalize the results based on your geographic affinity. E.g. if I search for "cardinals" then depending on my geographical distance to the center of the "cardinal" query location, my search results could be reordered to surface results that other people who live in the "cardinal zone" found helpful. No?

So this basically gives you another metric for user-similarity that can be used to do collaborative filtering of your search results.