April 17, 2008

Finding the Right Signals to Rank Search Results

Udi Manber, VP for search quality at Google, talks in an interesting interview about search, personalization and the influence of social networks on finding the information you need.
Search has always been about people. It's not an abstract thing. It's not a formula. It's about getting people what they need. The art of ranking is one of taking lots of signals and putting them together. Signals from your friends are better signals, stronger signals. On the other hand, many searches are long-tail kinds of searches. If you're looking for what movies to see tonight, your friend can probably give you the best information. If you're looking for the address of the business, the Web as a whole can give you better information. If you're looking for something obscure about anything, again the web can give you much better information. It depends on the type of search you do and how to take all those signals and put them together.

The ranking algorithm is a recipe that takes into account many factors: an absolute value for the importance of a page, the relevance of the web page to your query, the location connected to the web page, the recency, the percentage of people with similar interests that found the web page helpful, the density of information etc. It's important to find the right signals, their importance and the contexts where you should apply them, but Google relies on engineers to adjust the recipe. At some point, Google will have to come up with an algorithm that automatically identifies potential signals.

"I found this surprising. Google manually comes up with tweaks to its search engine that only apply to a small percentage of queries, tests the tweaks, and then tosses them into the relevance rank? (...) Frankly, I thought Google was beyond this. Rather than piling hack upon hack, I thought Google's relevance rank was a giant, self-optimizing system, constantly learning and testing to determine automatically what works best," wrote Greg Linden in "The perils of tweaking Google by hand".

{ via Slashdot }

1 comment:

  1. I am not sure whether Google was the first search engine that fine tune results through hand-picking.

    I heard that one of the search engine in China call QQ (started as a IM) is going to maintain its search engine through a team consist of thousands of people. Some other search engines in South Korea had similar practice in smaller scale.

    ReplyDelete

Note: Only a member of this blog may post a comment.