Re: [freenet-devl] freegle.com: major overhaul -> NeuroGrid

[Sam Joseph]

Chris Anderson Wrote > On Wed, 18 Apr 2001, Sam Joseph wrote: > > > > If the key index is being updated with new keys and being restored in > > Freenet couldn't you be updating degrees of association of keywords with > > keys based on user feedback? > > > > e.g. utdqti3tro7qiet6atsql2iugakfyg Douglas (6), Adams (5), > > Hitch-Hikers(4) > > > > I mean effectively you would be storing what NeuroGrid stores > > efficiently in a relational database in a flat file, but you could still > > do it, no? > > > > How are you incorporating this feedback into the rank? Suppose I had > another key: > > utdqti3tro7qiet6atsql2iugakfyg2 Quincy (12), Adams (10), Bush (8) > > If my query contains the term Adams, it seems the relative weights of the > term associations are the same, but the absolute strength of the Quincy > hit is stronger. Right, actually NeuroGrid uses a more complex ranking system where every rank is stored as a vector, so for example we might have: utdqti3tro7qiet6atsql2iugakfyg2 Quincy (12, 23, 45), Adams (10, 12, 25), Bush (8, 100, 1300) Where the ranks indicate the number of times a user has bookmarked something after searching for it with that keyword, the number of times it was clicked through after it was searched for using that keyword and the number of times it was returned as a search result for that keyword. You can then use some prob maths to compare the ranks. See: http://www.neurogrid.net/WhitePaper0_3.html For more details of the maths that can be used for this. CHEERS> SAM http://www.neurogrid.com _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://lists.freenetproject.org/mailman/listinfo/devl