Quarter Life Crisis

The world according to Sven-S. Porst

« NotesMainDelays »

Finding

720 words

Just the little rant on finding things on the internet that I promised in the previous post. Apparently, and without major contradiction, things like It’s more important to know how popular something is than to know what it’s about. can be claimed.

I think that's wrong. Surely it's more desirable to know what something is rather than only it's popularity. This would make refining search results much more efficient. This is also how well-greased search systems as seen at amazon, say, work: While they offer the most popular products first, they're also able to group them reasonably and enable you to refine your search accordingly. This can be very important if you start off with fairly general search terms that occur in many different areas – Google will return randomly mixed results which are often unrelated in that case.

Just take the query for my surname as an example. It will return some private home pages and a lot of items about a photo company with the same name, possibly having separate pages for their various branches, press clippings of how they went titsup &c. Wouldn't the most natural thing to group all those together and let people either exclude groups or restrict to them? Would a better understanding of what's in the index help here? I think so.

"But it's hard to do," people may scream. Well, maybe. I don't care. Surely, iTunes was hard to do, or TeX or Photoshop and so on. That useful things are hard to do, hardly ever kept people from doing them – unless they were Homer J. Simpson. The Google approach to web 'searching' may have taken us far but it is limited. This is obvious not only since people try to do some Google-spamming but even more so since the second superpower thing happening. Google's results actually help form opinions as people tend to trust it more and more. To establish something as 'fact' all you need is sufficient Google juice to push that bit to the top search results.

I saw this at work – in a non-dangerous way – the other day: There are quite a few people passing by here from Google for my Adam Green pages. So I was curious whether I may be on the first page of a Google search for 'Adam Green' – and I am. But in this case, Google actually works just fine as the most relevant site, the 'official' site is at the first position. However, on second position is a post from Scripting News mentioning someone else by the same name. Just being mentioned there without a link or other credentials given, makes it to position number 2 in Google, presumably because anything written at Scripting News is popular.

Now how can that be fixed? Fixed in the sense that search results are more predictable. A way to go seems to be using Bayesian filtering. I mean, man and dog are doing this for e-mail these days why not try to incorporate it into search engines? Then your Google search gave you a bucket full of links for Adam Green, the singer, another one for Adam Green the database guy; one for Photo Porst, the store chain, another one for me &c. Getting this to work may not be trivial, but it shouldn't be outright impossible.

Another hint that Google may have become too important for its own good is that Le Monde diplomatique has an article The world according to Google. This is remarkable because technology rarely makes it into Le Monde diplomatique. Not because they're technophobe, but because – no matter how excited we may become about new iPods or IE bugs – these things are rather irrelevant compared to things like the Middle East, Africa, freedom of speech, globalisation &c.

And if Google turn out to be actually good guys who strive to give the best search results rather than manipulating public opinion – then the reason why they may be perceived as bad is that their technology isn't good enough to be up for that job. And that in turn, in my opinion, is due to it's simplistic approach of relying too heavily on the popularity of sites. That was cool and new a few years ago but it won't do in the future.

October 14, 2003, 2:35

Comments

Comment by J. Scott Johnson: User icon

Hi,

I just wanted to say thank you for this post. While our opinions differ on some things, we’re far more alike than we are different and I completely respect your opinions.

Rest assured that I read (and pay attention) to pretty much all of your posts. Due to resource constraints we can’t really act on them as quickly as I’d like but I do hear you.

October 14, 2003, 18:31

Add your comment

« NotesMainDelays »

Comments on

Photos

Categories

Me

This page

Out & About

pinboard Links

People

Ego-Linking