Virtually!

Try searching for “Web Designer” on google. I did that yesterday as I was searching for a good freelancing designer on the web. It gave me the following results

Results
2 results - pointing to designers websites
Remaining - Rubbish

More than half of the results were not relevant for me. What other possible meaning could such a query mean other than looking for a “web designer”? A significant number of users would use that query with the expectation that the results would be a list of top web designers. I have been noticing this need repeating in many of my searches and not exactly limited to the above query.

So .. Is there a different way that we can get a system to understand what we really mean than to play around with ‘keywords’ to obtain the relevant results and fool the rankings?. Here is an attempt.

Consider del.icio.us. Its a community approach at organizing the web( in whatever form ). “Tags” have become the way to organize the web and tagging can enable a user to create keywords to remember a link with. So what pages would be tagged “web designer”?
Here is the result:
7 results - pointing to web designers website
2 results - pointing to resources for designers
1 result - pointing to something random

Voila! .. its so much better. But they still are not my top 10 lists but they definitely are more relavant. But the above approach has a significant limitation. It depends upon people to find the page and tag it else the page cannot be found. This cannot happen for the zillion other pages in the world.

Now, in reality different people tag pages in different ways with certain common keywords. In effect they give a personality to those pages by choosing a bunch of words. So what we end up with is a bunch of personalities for a page just like humans. This is similar to an individual’s take on what kind of “personality” another possesses. Each person’s take might differ and may have something in common.

This got me to start wondering about the possibilities of using an “Artificially intelligent” way to tag the web. But if you have a robot figuring out a way to tag information then it would end up associating a particular personality to a page which might be robotic in nature :). But then lets say we could somehow create different ontological descriptions for the same page with these robots. Apply page rank or Personality Rank (TM - I need to figure this out still ;)) relevance on this and I bet we will have better search results.

Now .. I will go figure out my robot and Personality Rank(TM). If you have any ideas please shoot :).