Monday, November 6, 2006

The search engine that could



Last week, Ethan Zuckerman wrote a great article explaining why Custom Search Engines would be so useful for communities such as the Global Voices network. He built a CSE to search over 3000 blogs from across the world. This is exactly the kind of application we built our platform for -- not just because of the scale of his search engine, but also the cause it serves and its collaborative approach. Unfortunately, as he explains in his article, queries on his search engine for some of the terms he's interested in didn't work very well.

After a week of some serious engineering, we believe we've made searching on Ethan's Custom Search Engine -- and all CSEs, for that matter -- much, much better. In particular, Ethan's search for Ghana, which originally returned only three results, now retrieves a much healthier number. All around, you should see much better performance in the quality of the search results.

The reason is fairly complex. Custom search engines are based on approximation algorithms that aim to search over the entire contents of the sites you specify. As with all approximations, there is always room for improvement. We're constantly working on our algorithms, and your search engines will continue to get better. If you see any anomalies or problems, please let us know. We want to hear from you about what is and is not working.