On October 21, 2004, Jeff Dean of Google gave a taped speech to the University of Washington’s Computer Science department. Lucky for us, the UW put this ~55 minute speech on some of the more technical aspects of how Google’s data retrieval model functions on the web.
Thanks much to Dominic of DigitalPoint forum for pointing this out.
The speech does not cover much on the subject of the search ranking algorithms, but it does talk about the spidering process somewhat and the indexing and retrieval procedures in depth. The most interesting part of the speech is the fairly detailed explanation of clustering technology, including an example of Google’s software that allows us to see the percentage chance that a particular term is connected to another.
The best thing about this speech, from the perspective of an SEO, is that it lets you think more like a search engineer. There’s not a lot there that’s completely over your head, which is nice, but there is a lot of information on the analysis of query popularity (full moon, eclipse, watermelon – around July 4th, etc). I urge you to spend 55 minutes watching the video during lunch (instead of reading that blog on American Idol).
I have excerpted the most important 15 minutes which are now available for download through SEOmoz:
Remember to use the right-click and “save target as…” Firefox in particular will not work with .wmv files if you left-click.
High Bandwidth – 78MB
Medium Bandwidth – 26MB
Low Bandwidth – 9.5MB
