Powerset has just launched. It's a search engine for Wikipedia that attempts to extract meaning out of the Wikipedia articles. I'm not convinced of its immediate usefulness, particularly given all the buzz it's gotten, but it's a cool application.
It works best when you aren't just looking for a particular Wikipedia article; for that, you can just use the site search within Wikipedia. It's more useful for subjects that span a number of articles, and for times when you want to quickly extract factual info (who, what, when, where) from a lengthy Wikipedia article.
So, for example, a search for Barack Obama turns up a moderately useful page with the basic facts (date of birth, names of family members) as well as links to all the Wikipeda articles that mention him.
Right below the basic facts is what are called Factz -- the most interesting part of this search engine. This is where the sense-making comes in. They are, according to the help file, "things" connected by a "relationship." So, for example, for Barack Obama we see:
attended: seminary, church, school, Occidental College, Islamic school
introduced: Iraq War De-Escalation Act of 2007, Response Act, Pandemic Preparedness, Transparency Act, and Federal Funding Accountability
and so on.
An interesting feature is that, instead of showing the usual snippets of text for each search result, you can get full sentences from each retrieved article. And there's a cool "miniviewer" that lets you preview the full Wikipedia article without leaving the search results page. (Kind of like Snap.com in that regard)
One of the reason it works at all is that it's dealing with standardized content and predictable fields.
I'm not sure when it would be all that useful, but it's a cool example of what kind of sense-making is possible.
Comments