Powerset hoped that its natural-language technology could create a new search engine that would let users type queries “in plain English, rather than using keywords,” according to a 2007 New York Times article. The startup raised $12.5 million but it only got around to launching a search engine that sifts through Wikipedia pages. Instead, it sold itself to Microsoft (NSDQ: MSFT) for $100 million.

ChaCha raised $16 million in 2007, offering the ability to communicate with a human guide while searching. The company switched strategies in early 2008, moving into the mobile search space.

And then there’s Amazon.com (NSDQ: AMZN) which launched a new version of its search engine A9.com in 2004, letting people personalize searches. The WSJ said at the time that it was “one of the strongest signs that it hopes to compete with the likes of Google Inc. and Yahoo (NSDQ: YHOO) Inc.” But Amazon scaled back the search engine in 2006 and it now exists mainly as a way to search for products.

WolframAlpha is arguably somewhat different. It’s not trying to offer a comprehensive way to search the web, but rather aims to answer “factual queries.” Search your name, for instance, and chances are WolframAlpha will give you nothing at all, or will try to compare two cities (in my case, Joseph, Oregon, and Tartaro, Bulacan, Philippines). But while its data-driven results give it a niche, they also limit its audience. Many of those who try it out will find it neat—but also may not really have a need to return.