[Search-l] Simpy Chichimichi - "Wikia Search - Not Happening"

Nathan Braun nathan at litepost.com
Sun Apr 29 19:04:04 UTC 2007


On 4/29/07, Stavros Macrakis <macrakis at alum.mit.edu> wrote:

> What's more, despite the community having been asked to contribute to the
> project, there is no channel for contributing. Consider my own situation...

Hey this seems as good a place to chime in as any?

I've been thinking about this problem for a while (as many of us have)
and have come up with some ideas I think could prove really useful.

I think one area people have not explored nearly enough is the true,
personalized contextualization - or 'hyperpersonalization' - of search
results.

To date, Search has generally been treated as "one massive problem" --
as in "one size fits all."  with one solution suitable to everybody:
that's Google.

But I don't think this is the right way to approach the problem at
all.  What people are searching for is highly individual, and we need
to treat it as such.

I think the core problem we need to address, if we want to build a
world-beating search application is: "the evolving matrix of
authority."

Let me explain what I mean by that (for those who are interested):

- By "evolving matrix of authority," I mean our informational needs
change over time.  This is sort of self-evident and obvious, but it
isn't sufficiently addressed by any existing information
infrastructure (ie including Google's Personalized Search and the
majority of other search ideas coming down the pipeline..)

The key problem is a way to auto-index or self-index, the "evolving
matrix of authority" for any individual person, in any individual
situation, for any given context.

This sounds difficult, but may actually be quite easy to do.

Has anyone seen the movie THE GAME?  In that movie, the protagonist
(played by Michael Douglas) goes through an intense barrage of
personality tests, in order to develop and design a Game suitable to
his needs.

I am suggesting that roughly the same thing needs to be done with
regards to search results.  There is hardly a "one size fits all"
solution.  Each solution needs to be individually tailored and
measured.

Fortunately, this should be relatively easy to do-- at least in
comparison to the Game's intense barrage of testing!

Search results could be improved enormously if each individual
searcher were to do a brief questionnaire prior to searching (or at
least after they had tired of the generic results).  Google is
probably experimenting with some sort of technology along these lines.

Such a questionnaire could simply ask (things like the following):

Which of the following is most authoritative for you?
[] The Bible
[] Wikipedia
[] Cosmopolitan
[] Google

... this could be a 2- to 4- to 10-page quiz.  The search algorithm
could be radically improved..for each individual user..on the basis of
such simple quizzing.

Alternatively, (ie in 'lieu of quiz') a persons Bookmark Folder could
be analyzed, to help come up with appropriate sources of information
FOR THAT INDIVIDUAL.

I'm not going to elaborate extensively here, but suffice to say, this
is one main avenue where traditional search falls severely short, and
where it could be radically improved.  I doubt any 'mainstream search
company' is going to take up tactics along these lines anytime soon.

People need to learn that there is not "one main authority" (ie
Google, Wikipedia, Cosmo or the Bible) but that authority varies by
individual taste and preference and belief system.  Search results
need to take this into account in order to be truly effective and
successful.

The best part of such a system, is that it could be cleverly built -
in a decentralized way - usefully integrating aggregating meta-data
from the existing search engines/results.  I mean, I can see how this
approach could literally be built "right on top" of Google, by way of
Wikia/Wikipedia (a perfect place to utilize such an approach).

I for one (probably like many on this list), often want to prioritize
Wiki results in my google results (as a result of "the evolving matrix
of authority" i refered to earlier, where Wikis have increasingly
greater and valid authority), and so I google:

whatever I'm looking for  wiki
(the very act of doing this indicates that wiki _for me_ generally has
greater authority than the unedited, raw Google search results
themselves.)

Google results are not "objective," as they claim to be - they are
highly subjective (determined in part by the subjective preferences -
and thus the collective accumulated experiences - of whoever created
the algorithms to begin with).

What search needs to do is become highly subjective -- ie
intersubjective and thus personalized.  Another word for this is
'hypercustomization' and it is the only way of the future, IMHO, as
far as search goes.  Everything else is just re-inventing the wheel -
er, Google.

etc etc ad nauseum

Please provide compelling evidence to the contrary so we can demolish
this line of though in the most systematic, scientific way possible.

Warmest Possible Regards,

nathan braun




The core idea I



> have some real experience in Web search, having been director of Search
> Innovation at Lycos (which no longer does its own Web crawl) and responsible
> for the MetaWeb Web mining project (which does) at Fast Search. At Lycos, we
> worked directly with engines like Ask, MSN, etc., and spent a lot of time
> listening to pitches from innovators. I got the Best Paper prize at SEM last
> year for my work at Lycos. I'm quite familiar with Wikipedia, having
> contributed thousands of edits and written tools to mine semantic relations
> from it at Lycos.  And I'm also an active contributor to the open-source
> Maxima project. So I am not a newcomer to Web search, to Wikis, or to open
> source.  What's more, I am ready, willing, and able to contribute to Wikia
> search. But I can't see any way to contribute to the Wikia search project in
> a meaningful way (and yes, I have asked).
>
> So all that is left for us on this list -- people who, as Aerik Sylvan says,
> are a priori supporters -- is to speculate.  Not the sort of community
> involvement I had in mind.
>
>                -s
>
>
> _______________________________________________
> Search-l mailing list
> Search-l at wikia.com
> http://lists.wikia.com/mailman/listinfo/search-l
> Change options or unsubscribe:
> http://lists.wikia.com/mailman/options/search-l
>



More information about the Search-l mailing list