Hey all,
Just got back from China (and I’m still up, which makes this about hour 23… ), and found out that the Frank Talk about MSN Search Video Andy Edmonds and I made for Robert Scoble on Channel 9 is now up! You can download the video or watch it (I recommend the download myself).
Good stuff, check it out!
Had to wait a Long time for that one!
Enormous amount of useful material, hope you don’t mind if I take some of your ideas and try them out in the appliance space? The neural net based on feedback is a wicked idea.
Because, after all, wouldn’t one want the Final Sort of your hit results be the Users selecting some document for some query? I don’t have to use a neural net to do that, so I have to do some research to see where neural nets can be used in the relevancy world to understand it.
To track what documents most matched what queries only takes an extra table with a many to many relationship.
What really left me hanging was all the talk about the older engines, the first class of engines, your alta vistas, lycos, etc. that features from those would come back. Heck I didn’t even pay attention to search engines until I started using Ask Jeeves! So what features
The video helped me to understand page rank some more - I used to just calculate links in - looks like there’s more I have to track. I guess I knew that but never quite understood it completely. Well, I still don’t.
I don’t get the whole links in links out bit - Isn’t this just saying that the site rank + page rank that has the link that goes into some site add to that sites rank + page rank based on some modifiers?
The only thing I can’t figure out by just calculating site/page rank based on the site/page rank of incoming links, is…. how and the heck do you calculate that! If I add to site A some rank number based on the rank of a site B, and the text of the link coming in (adding to a modifier that’s in the table that tracks queries and pages, the many to many), then - how do you know the Rank of Site B in the first place until all the sites in the graph have been traversed to discover any links into Site B!!!
So its almost like complete graph traversal is necessary to get the page rank… I think you said something like that - thanks, I think I have a better page ranking system figured out now
That’s key to me getting VC you know - that and just saying I have a neural network would be cool to say. Even if it didn’t really help any 
I liked the other guy - he added alot to the video.
Can’t wait to see more videos like this
Excellent article !!!!)))