Andrew Lentvorski([EMAIL PROTECTED])@Fri, Sep 14, 2007 at 01:05:27AM -0700:
> Bob La Quey wrote:
> >OK. So create a better search engine.
> 
> Non-sequitur.
> 
> The problem is not creating a better search engine.  Those dead
> pages will still exist.  The pages with nothing but Google-dreck
> will still exist.  Even your mythical "better" engine will have
> to deal with that.
 
I don't see the problem as being /caused/ by Google.  It is caused
by people trying to manipulate Google's search service.  That isn't
to say that Google didn't unintentionally create an environment
that encourages that behavior, but a better search engine might be
able to compensate.  In fact, seems to me that this is exactly how
Google came to prominence.  Everyone knew how to manipulate
AltaVista, Yahoo, etc... but Google was useful because it hadn't
been figured out, yet.

Taking that point of view, it may just be necessary for a new
"mythical, better" search engine to come around every few years.
It certainly couldn't hurt if those new tools were able to
compensate for the abundance of garbage to provide more useful
results.

> The solution (for the end user) is for Yahoo, Microsoft, and
> Google to all split the search engine about equally.  Then,
> specifically gaming one engine will be likely to drop your
> ranking on the other two.  Suddenly, Yahoo can exploit the
> Google-dreck to clean up their listings and the Google-dreck will
> go away.

I understand your reasoning, but it isn't realistic.  The end users
would have to use multiple search engines, at the same time the
search providers are trying to /distinguish/ themselves from the
other companies.  At some point, one *will* be more useful than the
others.  Expecting the worldwide user base to constantly change
search engines to manipulate the providers of the content, &/or the
search service is far removed from real possibility.

> What I really want from a search engine is the "Junk" button from
> Thunderbird which is used to help train your spam filter.  When I
> run a search, I want to be able to classify sites as "Junk" so
> that they start dropping in Googlerank for me.
 
Bayesian filtering of search results -- now this idea sounds
useful.  It puts power in the users' hands, and would make it
/much/ more difficult for any content provider to even /guess/ how
search results are ranked for any given user, regardless of the
choice of search engine.

In the past, I was aware of a search provider called dogpile.com.
Their strategy was to query several of the popular search engines,
and provide a certain number of results from each of them.  It
sounded useful, but the presentation of the data left something to
be desired.

Let's say, though, that a user could control an aggregate search,
and then a bayesian type filter could re-rank, and collate the
results.  Of course, now we're back to creating that "mythical,
better" tool :-)

> Even better would be to be able to download and then share this
> with somebody else.  For example, I'm pretty sure that I would be
> very happy to add Stewart's "Google Junk Corpus" to my own.

I don't know.  On the surface this sounds good.  In practice, I
suspect that people will get better results by sticking with their
own data.  One concern is that, if one corpus of data became
popular, the content providers might find it useful to manipulate
the results at that level.  I might even balk at having a default
corpus for new installations of such a tool.  In the end, though,
the results for those who maintain their own filters would still be
much improved.
 
> >Meanwhile we muddle along with less than perfect tools, blaming
> >them if we must, but rarely throwing them away.
> 
> I don't think I agree with that.
> 
> I would argue that computer folks are a little too eager to throw
> things completely away when a little more polish is actually
> called for.

The problem isn't the tools, their quality, or their polish.  The
problem is human behavior.  It is generally believed among computer
folks, that tools can help avoid some of the results of that bad
behavior.  The more "mythical" the solution sounds, the more
interesting it, and maybe the results, are.

It's difficult to stay on top of changes in human behavior, and
examine new approaches, by only polishing the old tools.

Wade Curry
syntaxman


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to