Monday, May 18, 2009

Wolfram Alpha...something new

I haven't experimented with Wolfram Alpha yet.

"Wolfram/Alpha Looks Beyond Search Toward Computation"

Unlike Google, the newly debuted "computational knowledge engine" attempts to compute answers to questions based on established facts.


Thomas Claburn

May 18th, 2009


Wolfram Research, makers of the noted computational software Mathematica, on Monday launched Wolfram/Alpha, a service the company characterizes as a "computational knowledge engine."

It's not a search engine. Though widely compared to Google, Wolfram/Alpha does not try to match queries against a list of indexed documents. It attempts to compute answers to questions based on established facts.

Whereas Google takes an ambiguous query like "bass" and returns results related to the fish, the musical instrument, and the shoe brand, Wolfram/Alpha returns an array of nutritional data related to the fish.

Google can tell you where to find information about "the best restaurant in San Francisco." The data sources identified by Google may not be ones you'd agree with, but that's what you get when you ask a subjective question.

Wolfram/Alpha aspires to offer something more objective, answering queries not by matching keywords in an index but through computation. For example, if presented with the query $250 + 15%, it returns the correct answer, $287.50.

Google too can do this, if the query is constructed properly: (15% of 250) + 250. But Wolfram/Alpha is far more sophisticated in the kinds calculations it can perform, assuming it can turn the query into a proper equation.

When presented with questions related to its curated knowledge base that can be parsed, understood, and answered through computation, it performs brilliantly. Asked to find "the distance between New York and San Francisco," it not only returns 2,578 miles, but it also provides a calculation of travel time for a person in an airplane traveling at 550 mph (4.7 hours) and for light traveling through a fiber-optic cable (19.4 milliseconds), among other fact-based calculations.

Google returns links to pages that contain the distance between New York and San Francisco, but lacks the travel time computation. Where Wolfram/Alpha shines is when presented with queries like "distance between Mars and Saturn." It returns what, presumably, is the correct current distance between the two moving planets at the moment, 10.79 AU, or 89.76 minutes at the speed of light.

Google isn't even close: The first document returned is an article on that mentions Mars, Saturn, and Saturn's distance from the sun. But it makes no mention of the present distance between Mars and Saturn. And the second document returned is a page that erroneously indicates the distance between the two planets is 23 miles.

Stephen Wolfram, the founder and CEO of Wolfram Research, describes Wolfram/Alpha as a "new paradigm for using computers and the Web."

Indeed, it will make answering certain computable queries far easier than before. But Wolfram/Alpha is likely to remain a research tool for relatively sophisticated users until its knowledge base expands and its ability to understand poorly crafted queries improves.

At a recent media event about the state of search technology at Google, Google engineering VP Udi Manber observed that "if users can't spell, that's our problem." Google deals with misspellings and other query problems and its users appreciate Google's efforts to understand them.

Wolfram/Alpha does some spelling correction but it's far less tolerant of bad user input than Google. It's a bit like a child prodigy: brilliant in some respects, but lacking in the social graces necessary for broad acceptance. As it grows up and develops, its power and utility will become more widely apparent.

The Geek’s Search Engine?

No comments: