Division Trust
Wednesday, December 28th, 2005
Line-making and Line-tending in Light of Trustrank
The classification systems we use everyday and all over the world are based almost entirely on arbitrary and subjective differences we decide exist between things. That two people can disagree on whether Object A belongs in Category 1 or Category 2 begins to illuminate an important problem faced by the Google Spider, as it is today. If two persons, with all five senses in good working order, can disagree on fundamental classification, how can the Spider, which is blind, serve them both?
To remember Image Search in its beginnings, you’ll recall SERPS filled with very generic looking, you might even say idealized, images of things. That was when the Spider was young and wholly literal, operating on the basis of simple text recognition for a pre-programmed list of words. If it encountered a file named apple.jpg, it assumed “apple” — which is exactly what its programmers wanted it to do. Programmers, who were also responsible for the initial “seed set” of images, chose ideal representations of basic objects, like the perfect red apple, a tree set by itself on a grassy hill, a pen on a desk, etc. Most people would probably do as they did, if given the job of “teaching the meaning of things” to a fledgling intelligence — read more about Division Trust and Trustrank.
google - trustrank - blogging - porn - profitlab - mardack - trustrank