Google Suggests Guts Disassembled - Part 2
Slashdot introduces us to this analysis of Google Suggest, which goes even deeper than previous dissections of the Google Suggest engine. Some of the interesting discoveries:

A simple program which finds all possible suggestions for any given starting term.
Google Suggest ignores quotes.
Order is more important than the actual terms.
The number of suggestions is actually smaller than expected, making it possible for someone to implement Google Suggest on their own server, even implementing the whole thing in RAM (its that small).

The most startling thing is that Google Suggest is actually based more on searches than results. To explain: Google Suggest returns results that are not in Google's index, or for terms that Google can never get to, because it indexes searches made as well as searches found. What does this mean? If you have typed a UPS tracking number into Google (something typical, because Google has searches for tracking numbers built it), it can find its way into Google Suggest. Just go there and type in "1ze" and watch the numbers pop up (all from packages delivered in the last six weeks). Does this mean credit card numbers could be in there as well? Less likely, but possible. Ironically, if you've ever searched for your credit card number to make sure it wasn't publicly available, you may have inadverantly added it to Google Suggest. Oy.

Related posts:
Google Suggest - 12/10
Google Suggests Goooooooooooooooogle - 12/10
Google Suggest Tools - 12/11
The Google Suggest Complete My Sentence Game - 12/15
Google Suggests Guts Disassembled - 12/18
Google Suggest Poetry Generator - 12/20

- posted by Nathan Weinberg @ 12/24/2004 02:34:00 PM