Search Tech

Number-Crunching Search Engine WolframAlpha Preps for Launch

A new search service that brings a measure of artificial intelligenceto traditional Internet searching is set to go public soon. Wolfram|Alpha has been anticipated for several months, and the company behind it recently drew attention by opening access to a small number of testers.

The service should launch late this week, states the company’s blog, though it names May 18 as the official date.

The brainchild of creator Stephen Wolfram, Wolfram|Alpha is acomputational knowledge engine that analyzes the relationships amongsearch terms to create an answer to a search query. It is based onWolfram’s math formula software Mathematica, a tool for scientificcomputation that also functions as a programming language.

Wolfram Research, the company Wolfram created, has been providingselective access to testers in preparation for the planned launch.

The Wolfram search engine is not likely to go into head-to-head competition with the likes ofGoogle, Yahoo or Microsoft Live Search. Instead, it will provide researchers withan alternative approach to traditional searching.

“What they are doing is brilliant. This will shorten people’s worktime. It is a great boon for productivity,” Brooke Aker, CEO ofsemantic search technology firm Expert Systems, told TechNewsWorld.

What It Does

Wolfram|Alpha is not a traditional search engine that plows throughWeb pages based on keywords. Instead, it computes a response based onthe data contained in the company’s own knowledge database inconjunction with an artificial intelligence system.

What sets Wolfram|Alpha apart from other search engines is itsability to do sophisticated pure computations involving numbers orformulas users enter. These computations are applied automatically todata called up from its repositories. Computation is what turnsgeneric information into specific answers, Wolfram wrote in a May 1 blog entry.

Traditional search engines like Google and Yahoo crawl the Internetwith spider programs to present a catalog of Web site content.Wolfram|Alpha, on the other hand, attempts to answer natural language questions, explainedScott Testa, professor of marketing at St. Joseph’s University.

“This takes semantic search to the next level. Natural language search … is the Holy Grail of the industry,” Testa told TechNewsWorld.

Problem Solver

Wolfram/Alpha addresses a longstanding dilemma.

Search engines answer questions differently than people do inconversations, according to Nathan Myhrvold, founder and CEO ofinvention firm Intellectual Ventures. Myhrvold was a key player in founding Microsoft Research and numerous technology groups associated with some of Microsoft’s most successful products.

“Stephen has developed a system that answers natural languagequestions. This is definitely brand new technology. People have beentrying natural language searches for a long time,” Myhrvold toldTechNewsWorld.

The computer has to understand what the person making the query means,not just what the words mean, he added.

Different Approach

Wolfram|Alpha goes beyond common semantic search techniques. Most semantic searchengines attempt to interpret the search terms, but Wolfram’s method takesthe concept to a new level, according to Aker.

“I think the tools we have now don’t go far enough,” he said.

One example Wolfram offered in his blog: a computational search focused on finding the number of calories in a recipe. The basic data involves the calories per gram ofeach of the ingredients. However, turning that generic information into theactual total calories for a specific recipe requires computation, hewrote.

For instance, unit conversions of so many cups into grams of flour andthe default weight of a standard egg have to be computed to multiplythe calories per ingredient and then totaled.

Mathematica Basics

The math formula system Wolfram developed over the last 20 years isthe foundation of his new search engine. Mathematica’s symboliclanguage lets searchers express complex computational processes in afluid, intuitive way. Users do not have to toil with what he called”the ugly details of data structures, memory allocation, or confusingand inconsistent subroutine libraries.”

The symbolic nature of the Mathematica language makes possible a great degree of interoperability between different parts ofthe system and between different algorithms and data sources. Itrepresents all kinds of data with arbitrarily structured symbolicexpressions, according to Wolfram.

The combination of Mathematica and the use of the highly specializeddatabase are what make Wolfram|Alpha so unique, said Testa.

“Every 12 to 18 months, a new attempt surfaces to beat Google. Mostfade away. Wolfram is playing it low-key … . My gut is that his existing reputationis going to make sure that this is something pretty significant,” he predicted.

Unanswered Questions

The real significance of Wolfram’s new technology is the impact itwill have on the targeted users. This certain user class can get exactanswers. It is a more efficient allocation of resources, said Expert Systems’ Aker.

However, Wolfram Research so far has not spelled out exactly who thoseusers are and how much, if anything, the service will cost. For instance, generalusers may be provided with limited access for free. Professional users might be offered more features for a subscription.

Another unanswered question is whether the company’sproprietary database is robust enough. In all likelihood, some questions posed in thesearch window will come back blank if there is not enough data compiled.

“Lycos, the first Internet search engine, was awful when it started.Even Google was not as good when it started as it is today,” Intellectual Ventures’ Myhrvold pointed out.Wolfram|Alpha will find stuff that you could never find with Googletoday. I think people will be excited about the launch and will findAlpha useful. It will have the potential for some users to becomefrustrated.”

Too Limited?

Some criticism of Wolfram’s approach could focus on the end-userapplication of these search concepts and tools. Key areas of focus for thesearch engine include Machine Learning, Semantics/Natural Language Processing(NLP) and Common Sense Reasoning. The key is to proactively bring thecontent that matters most to the user, as well as connect them withothers who are looking for the same thing, according to Jim Anderson, CTO ofSaber Seven.

Saber Seven is developing a solution that brings the answers to theuser and refines the process with all these techniques integrated intothe experience.

“Wolfram Alpha might be helpful for confined, factual questions like’What type of metal is the Liberty Bell made of?'” Anderson told TechNewsWorld. “However, thesesimple and very specific questions don’t address the real, complexquestions people are looking for, like ‘I’m writing a paper onfreedom. Where do I start?'”

1 Comment

  • You clearly nailed down the difference between Alpha and Google. It is not meant to replace general search engine—rather a new kind of SE.

    Likewise, at Cazoodle, we are building vertical search engines in specialized domains, such as Apartment Rentals, Local Events, and Online Shopping. Clearly, these do not exist today—and are not meant to replace general search engines.

Leave a Comment

Please sign in to post or reply to a comment. New users create a free account.

More by Jack M. Germain
More in Search Tech

Technewsworld Channels