Thursday, December 01, 2005
Capturing Local ContentI got a little bit of a late start this morning so I missed the big Google keynote. Greg and everyone else I spoke with said the keynote was awesome and such so... if you are at ILM 2005 and you went to the keynote then please submit your thoughts on the talk in the comments of this post.
I decided to go to the capturing local content session instead of the advertising session because paying for advertising is kind of a no-brainer. You've got pay per call, pay per click, pay per acquisition and cpm... the list goes on but you get the idea. This session however is about adding structure to local search content and it should be interesting. The list of panelists goes like this...
Michael Adelberg: Strategic Partner Development Manager at Google
Stephen Baker: VP Portals & Directories at Fast Search & Transfer
Matthew Berk: CTO Openlist [Josh Stylman says I should get to know this guy]
Greg Gershman: President Blogdigger [killer blog search engine run by two guys with full time jobs - can you say bootstrapping 101?]
Andy Moss: AWOL
Azim Tejani: COO i411 Inc
So the first question is about search relevance in local search results. Tejani says it's impossible to see who's relevance in local search when you're looking for a local plumber. I firmly disagree with what he's saying. Vertical search is all about tweaking algorithms around a specific data set and user mindset.
Google's Michael Adelberg says they do try to apply some logic via multiple algorithms to their local search product. Admittedly their logic uses formulas and such developed by engineers so it might not be the right method for everyone but they are trying to determine what the user is searching for.
Greg from Blogdigger posed the question about whether or not there is any human input on the relevance of results and Adelberg said
"the algorithms are written by humans so yes there is a human element but it's the computers that are doing the work..."A theme that's coming out is that the relevance of a search result set is based on what the underlying content is rather than the context of that content. I bet Ray Kurzweil would have a thing or two to say about this issue in search, can you say AI?
Stephen Baker said we need to move from text mining [crawl based results] to fact finding. The key difference here is how much logic is applied to your crawler's filtering so that results that aren't relevant are never even included as part of the index.
By Jason Dowdell at 11:12 AM | Comments (0)