Slashdot Mirror
MS Patent Applications Reveal Search Technology
eldavojohn writes, "In the roughly 90 patents they applied for on November 2, 2006, Microsoft reveals that it is apparently pushing its research in the search engine market. There are a few patents that reveal improved ranking methods and document classification but the real interesting ones revolve around linking related queries, optimizing search, identifying results that are spam, and using a Bayesian classifier to measure feedback from the user. If that's not enough, there's even a few I don't quite understand. Another notable Microsoft application for a patent is the model for assisting children in authoring stories so you can't accuse Microsoft of not thinking of the children. Microsoft regularly applies for many patents but never so many revolving around search."
This patent sounds like they've patented the idea of recording which search links are clicked on most often for a given query, thus providing feedback given a random session number of which links you clicked on, and if you came back and tried something else.
it would improve search results for future searchers, but I dunno if I like the idea of my search being tracked.
Not only do my searched probably already get tracked without my knowledge but I might be completely wrong about this patent as I only perused it.
CAn'T CompreHend SARcaSm?
Comment removed based on user account deletion
A hit is a hit is a hit, regardless of the intent. You might not like it but the numbers aren't false.
.com at the end" and ignores the results? Or better yet, how many hits resulted in an intentional search? *That* would be a more useful number. I've always had a problem with how folks quantify how well a thing is on the Internet by the number of hits it gets. Hits have the potential of being arbitrary and artificially inflated, as the original poster stated, and in my opinion provide little useful information.
/gam/
That may be. But how many non-false hits were unintentional and thus serving useless search information to someone that just says "whoops, I meant to type in the
"In theory, theory and practice are the same; in practice, they are not."
This Patent application for a system to analyze and compare of portfolios by citation submitted by Microsoft sounds like it might actually help a patent examiner find prior art for all these Microsoft software patents. It describes a system for classifying documents and finding and analyzing relations (citations) between two sets of documents. Although that does sound a lot like PageRank and anchor text analysis doesn't it?