[Search-l] NPOV for Search?

[Bryan Bishop]

On Tuesday 08 January 2008, John McCormac wrote:
> The only mechanism that works in an repeatable fashion is the use of
> words in the query. Trying to move beyond that is difficult because
> you are in the position of having to try to guess what the user

Yeah, this guessing game is hard without enough information. If the user 
provides the necessary context, this information can be capitalized on.

> wants. With search engines, you can only please all of the people
> some of the time. Google has a greater coverage of "some" than others
> like Microsoft or Yahoo.

I wonder why that is.

> > "Relevance" is not a concept that can stand alone.  Relevant to
> > whom, and for what?  Once we scratch the surface, we understand
> > that relevance asks for balance, for neutrality, for quality.

Re: quality. Quality is qualified, not quantified. Not universally 
quantified, at least.

> "Relevance" for the moment, is that which satisfies most users for
> that particular query. Basically a search engine has to provide what

And it can be much more in the future.

> the operators hope is the most likely set of results that are
> relevant to the user's query. Social search offers an element of

I wonder how Google has come out on top anyway, by pure link analysis? 
Maybe they do multiple crawls from different seed sites and that's how 
they are avoiding biasing their index too much.

> refinement but the hard work has to be done on the backend. This

Sort of. At the moment, social-net users are filling out tons of data on 
their own, right into the proper format that the websites want them to 
put it into. :) Hurray. So if we can somehow provide a one-to-one 
correspondence of user data input into changed search results ... there 
will not be too much backend stuff going on. But again, see my other 
email about dropping in developer 'scoring' plugins or something. 

> might be putting the cart before the horse at this stage but the
> quality of the search engine's URL set has to be clean and usable.

Right. That's definitely a problem. John, you have been informative in 
IRC concerning how the domain names are tainted and ownership switches 
and how this taints SE indices. If somebody else agrees with me, I 
think you should send an email to the list. :)

> The neutrality aspect is what concerns me most because every
> successful search engine is far from neutral. Perhaps it is just a
> philosophical argument rather than a technological one and I'm
> thinking as a search engine techie.

Nah, it's a tech problem. Special niches of developers can come in and 
populate the landscape with innovative ways to search *from particular 
contexts* of the users [that they develop for]. And I don't think I 
have to reiterate the business aspects re: colonizing parts of 
MPOV-space first. :)

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