Clever search goes live

It has been tantalising the internet audience for a while now: the intimation of a clever search that understands humans and gives results that are not just more or less relevant, but actually represent a correct answer to a search query.

In May 2009 a new “knowledge engine” goes live. Named Wolfram Alpha, it claims to have the answers on (more or less) any question users may ask. The ambition of Wolfram Alpha is to combine natural language with the scrupulously collected information and a new search model to serve — the answer itself.

In other words, Wolfram Alpha does not simply return documents that may or may not contain the answers to your search term, like standard search engines do. It is not just a giant database of knowledge, like Wikipedia. It also does not simply parse natural language and use it to retrieve (hopefully, relevant) documents, like semantic web is doing (or trying to do).

So, what kind of questions will Wolfram Alpha answer? Here are some examples: “What country is Timbuktu in?”; “How many protons are in a hydrogen atom?”; “What is the boiling point of carbon dioxide?” and so on. All in all, Wolfram Alpha understands and then computes answers to certain kinds of questions from different aspects of human knowledge. The knowledge is collected by a team of human experts and aims to be 100% accurate. That knowledge is brought into a common well and (this is where the potential strength of Wolfram Alpha lies) made to interact, blending various bits and pieces of knowledge fed into it in order to give an accurate answer to a query.

Language being a tricky tool, the phrasing of the question can influence the results, as a New Scientist writer who tested Wolfram Alpha found out when trying to get the clever search engine to tell him what would be current value of $25 million from 1945.