Search is a highly specialized field, however billions of people interact with it as users. Then lay on top of that the huge business implications of the industry and you’ll see why there are tons of articles about search deals, the stock prices, and market share. All of these are important, but I’m always disappointed that the mainstream media generally ignore the actual products. Even search “experts” rarely write about the quality of search results and the technology behind the curtain. It’s so much easier to write about deals and market share than to do product evaluations. Other specialized fields, such as ichthyology and crystallography aren’t subjected to the mainstream media except when something big happens. Otherwise their literature is restricted to a limited audience and a limited group of researchers and writers. Not so for search.
This Businessweek article, A Gaggle of Google Wannabes, gets closer than most articles. In it the author actully gives a brief description of how Ask’s algorithms differs from Google’s. (Note: when I say brief, I mean brief, but still better than nothing). To quote: “Ask.com’s algorithm, on the other hand, retrieves and ranks results based on the number of times groups identified as related to the topic reference the site. Company executives say the method is superior because it theoretically avoids displaying generally popular sites that are not frequently referenced by other sites on the topic.” In this excerpt there’s a brief explanation of what’s different but then the article relies on “company executives” to say their product is superior. But nonetheless I actually did learn something about Ask.
UPDATE 11/7/06: The search engine that would outdo Google, by Bambi Francisco, is an article that actually does some quick and dirty relevancy testing of queries. Wow! She evaluates how Powerset compares to Google for certain natural language queries, and points out a couple of examples of how the two engines handle some stopwords and proximity issues. According to her evaluation Powerset does better with natural language queries.