The launch of Wolfram|Alpha in May was preceded by a huge amount of hype. Wolfram|Alpha was hailed as a potential Google killer and an entirely new kind of search engine. After a couple of months of being used by real searchers for real life queries it is clear that it is neither.

Wolfram|Alpha is not a Google killer. This is eminently clear if you take a look at the numbers. Andy Beal reports that Wolfram|Alpha’s traffic has dropped 82% since its launch.

But the verdict is supported by more than numbers. Wolfram|Alpha is in fact not a search engine at all. It’s a fact engine. Most of the queries you take to Google, Yahoo or Bing makes no sense to Wolfram|Alpha.

The input problem

Wolfram|Alpha is not good on natural language processing. In fact, it demands input according to quite particular rules. In a recent blog post about recent developments of Wolfram|Alpha Stephen Wolfram writes:

Close to half the time that Wolfram|Alpha doesn’t give a result, it’s not because it doesn’t have the necessary knowledge, or can’t do the necessary computation. It’s because it doesn’t understand what’s being asked.

The growing team of developers work on continuous improvements and has reduced the “fall-through rate” of queries we don’t understand by 10%.

Since the launch, 3907 that the bugs have been fixed. These fixes will probably make Wolfram|Alpha easier to use, but will it make the results easier to understand?

The problem with facts

What Wolfram|Alpha does is scour the Web for systematic knowledge and make these facts searchable. But presenting facts without any context is problematic. All facts require interpretation and Wolfram|Alpha rips the facts from their original context and offers no help in the interpretation at all.

And where are the facts derived from? In order to understand numbers and facts, it is important to know the source. And all too often, Wolfram|Alpha reveals no source.

Also, many of the “facts” of this world are contested. Scientific schools battle for their take on the truth and new discoveries are made all the time that render yesterday’s facts dubious.

Further development will hopefully improve the input process and this will make Wolfram|Alpha a good source for information about simple data like currency conversion, calendar data and geographic information.

For other, more complex queries, Wolfram|Alpha will be of little interest.

For one thing, writing queries will continue to be complicated — too complicated for most people. And even more problematic: The data delivered, presented as “facts” will be hard to trust, given the reasons presented above.

Related articles

Wolfram Alpha is the new Wolfram Alpha, not the new Google
http://www.pandia.com/sew/1824-wolfram-solves-problems.html