{loadposition jake}When Wolfram Alpha debuted last May, observers weren't sure what to make of it.

What it did -- computing results based on data on the Web and presenting the results in a graph or table format -- it did very well. But its usefulness seemed limited, and the task of writing a query to successfully get the answer you wanted could be daunting.

In a new blog post titled "What We’ve Been Doing This Summer", Stephen Wolfram acknowledges the difficulty: "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," he writes.

A big part of the effort this summer, he says, has been to develop techniques by which Wolfram Alpha can learn to interpret queries based on what it "reads."

Wolfram claims that the process has already reduced the comprehension failures by 10%.

Wolfram doesn't discuss it, but according to Internet business blog TechCrunch, the company has signed an agreement with Microsoft to supply data to its upstart search engine Bing.

Bing does a good job with extracting useful results, based on a searcher's presumed intent, and presenting them accessibly. That strength could make it a good match for Wolfram Alpha's weaknesses.

]]>