Wolfram Research’s new product Alpha to compete with Google and Wikipedia

This is the stable version, checked on 8 January 2018. Template changes await review.

Monday, March 9, 2009

Wolfram Research Inc., makers of Mathematica and A New Kind of Science, have released a limited alpha of their new web service Wolfram Alpha. Wolfram Alpha is described by CEO Stephen Wolfram as a "computational knowledge engine." Unlike Google, Yahoo! and other traditional web search engines that rely on statistical methods for retrieving online documents, Wolfram Alpha answers factual questions in the way that Wikipedia does, except it relies on analytical methods instead of human-generated documents.

Although the product is still in a limited release some details about its operation and design have been revealed by Wolfram on his blog, and by Nova Spivack who recently interviewed Wolfram about Wolfram Alpha. The product is available on the Web as a single search box reminiscent of Google’s main search page. The search queries can be entered in natural language and the natural language system will parse the query and use models of knowledge (ontologies) and human-curated data to return an answer including graphs and other representations. The ontologies and data are managed by Wolfram employees who must input new ontologies and data by hand or, occasionally, with the assistance of programmatic tools. It is expected that an Application Programming Interface (API) will eventually be available, although it is not known what the API will be used for.

Unlike traditional search engines Wolfram Alpha does not search online documents, and thus does not return answers to "fuzzy" questions, such as opinion or advice. Instead, the scope of answers is limited to the knowledge that has already been modeled and encoded in the ontologies and the associated data. In this way the system can generate knowledge that was previously unknown. It is thought that while Wolfram will concentrate on scientific and technical information the system may eventually be able to answer questions in other domains, such as stock information, geography and history. In theory, any question with a factual answer could be answered by Wolfram Alpha.


Sources