How to improve relevance

Filed under: Internet · Date: Sat May 17 10:00:05 2008

Google created a killer app for search engines by introducing relevance calculations which took linking into account. Sites which got a high number of incoming links were judged more relevant than those which had few or no incoming links.

Link relevance lasted so long until SEO firms found a way to abuse the system by creating link farms. The rel-nofollow tag was an attempt to fight back this, but since it was voluntary, it wasn't very good at it. Another point agains rel-nofollow is, that it was often automatically added for all comments or "foreign" content, thus preventing the web community from improving content or content relations.

Soon after Google picked up blogs, and started weighting blog posts higher. Blogs soon dominated the search. When bloggers found this out, they did some interesting experiments where they picked up a random word, linked to each others blogs with the keyword, and after a couple of days googled for it and found that their phony page was a top search result.

That was not an issue, until the same SEO firms that polluted the web with linkfarms, figured that blogs could be used for the same purpose the linkfarms served. And they could make an extra buck with Adsense too.

This all brings us to a situation where search engine results are not as good as what they used to be. So how to improve the situation? Here's my thought for a simple solution.

Improvement

Google monopolizes much of the webspace, and it could harvest huge amounts of personal data if it wanted to. However, this data can be put to a good use.

Googles Adsense tracks the time visitors spend in each page. The more time the visitor spends reading a page, the more interesting the contents must be. After all, who would spend lots of time on pages which they do not read? Using this logic would enable improving search result quality tremendously.

There are problems with this, of course, like spoofing Analytics transactions, but they can be easily combatted, if not done so already. I haven't read their Javascript code to see if it is protected against fake clients.


Comments are disabled for this post.