[Search-l] [canonizers] Re: NPOV for Search?

[Bryan Bishop]

On Wednesday 09 January 2008, Brent Allsop wrote:
> Yea, that sounds like a good idea.  Let me see if I'm following what
> you're saying.  So perhaps you could use the quantifiable POV
> canonized scores and rankings of web pages for search and stuff?  And
> since the scores would be based on who you choose to trust, based on
> what canonizer you have selected, you could better get what you want
> from the search since higher scored pages, based on your selected
> MPOV people (or just blind popularity for NPOV), would be given
> priority?

Somewhat. Jimmy mentioned in another email that he suspects this method 
of relying on a person's social network will be a bad idea. I think I'd 
have to agree. First of all, it would not be easy to change your 
network if you suddenly wanted to take up the MPOV of a researcher, or 
of a spy if you wanted to suddenly choose that. You wouldn't be able to 
go get those friends and stuff, and that's just not right.

Instead, we need some way to qualify the user, to describe the user, 
most likely provided by the user's own input. The challenge is to be 
able to make a correspondence between a user's context somewhat 
conveyed through options they select on a web page (plus feedback) and 
into the ranking algorithm of their choice. 

The user opinion data can probably be structured into XML. Check out the 
dataportability.org group. Somehow, developers will have to figure out 
how to mine the provided XML descriptors on each user and apply it to 
the MPOV SE.

- Bryan
________________________________________
Bryan Bishop
http://heybryan.org/