Sponsored By

Microsoft Buys Search Start-Up Powerset

Under Powerset's hood is decades-in-the-making natural-language processing technology the company licensed from the Palo Alto Research Center, formerly of Xerox.

J. Nicholas Hoover

July 1, 2008

4 Min Read

Microsoft's proving that it has plenty of ways of spending the $40-plus billion that Yahoo turned down earlier this year, announcing Tuesday that it would buy search startup Powerset for an undisclosed sum.

Powerset brings Microsoft a new line of attack against Google, and one that doesn't rely on quickly gained economies of scale like an acquisition of Yahoo would. Instead, Powerset brings new technology to Microsoft: semantic search.

Semantic or natural-language search relies on sentence structure, syntax, dictionaries, and thesauri to extract meaning from text, rather than relying on how heavily Web pages are linked to one another to determine the relevance of search results. Under Powerset's hood is decades-in-the-making natural-language processing technology the company licensed from the Palo Alto Research Center, formerly of Xerox.

"Powerset set out with a really ambitious vision to bring about the future of search through machines that could understand language," Powerset co-founder and CTO Barney Pell said in an interview. "One of the questions is, how is a company that's a small little startup like Powerset actually going to bring that out to a large scale? With Microsoft, we've found the right partner. The vision, the goals, and even the culture were really aligned."

Though today's Web search companies employ linguists and can distinguish between singular and plural words, correct some misspellings, and find some synonyms, they're nothing like search engines designed from the ground up with contextual meaning in mind. Powerset competes with Hakia, Cognition Technologies, and several others.

Microsoft estimates that one-third of all searches don't get answered immediately, often because the search engine can't pull meaning out of queries. "To go to the next level on search, we can't treat search as matching keywords and queries," Ramez Naam, group program manager for Microsoft Live Search, said in an interview. "We have to go into understanding intent, meaning, and context in the queries so we can do matching and ranking at that level."

For Powerset's startup competitors, Microsoft's acquisition will likely be seen both as a validation and a warning shot. On the one hand, there's the view of Melek Pulatkonak, Hakia's chief operating officer. "This is definitely a confirmation and a vote of confidence that semantic search is here," she said in an interview. However, if Microsoft's acquisition is the beginning of a chain reaction, winners may be determined more by who gets acquired than by who has the best technology.

Google, for one, isn't saying much. "Search is a highly competitive industry and we welcome competition that stimulates innovation and provides users with more choice," the company said in a generic-sounding statement Tuesday afternoon. There are still doubts about the efficacy of semantic search. Language isn't something as easily dissected as mathematical linking algorithms. Many searches get better responses from Google than Powerset, though that's understandable because Powerset's public site currently only searches Wikipedia. And Hakia's Pulatkonak, for one, cautions that Powerset's technology relies too heavily on decoding syntax and not heavily enough on parsing meaning from ontologies, or relationships between concepts.

Pell admitted Powerset hasn't solved every problem, but Microsoft gives the company a big boost of manpower and research dollars. "Powerset has been trying to draw on the best approaches we could," he said. "With the resources combining, we'll be reaching out even more as we bring these things forward. It's not about one specific technical approach or another; I think it's about bringing all the best approaches together."

Those gains may start coming soon, as both Powerset and Microsoft say there are "many" announcements to come. Expect Microsoft to put Powerset's technology into place at several junctions, potentially beginning with specialized site searches like Powerset has already implemented with Wikipedia. Eventually, Powerset's technology will be embedded even into generic Web search. "You're going to see Powerset technology affect the relevance of those 10 blue links," Naam said.

If Microsoft is to include Powerset's technology in its relevancy calculations, it also will have to take into account semantics during its indexing process and will require significant testing and jimmying of its current algorithms to correctly balance the weight of semantics and traditional ranking techniques.

Terms of Microsoft's deal to acquire Powerset were not disclosed, but Venture Beat reported last month that the acquisition was for slightly more than $100 million. Rumors of an impending deal have been circulating for several months.

About the Author(s)

J. Nicholas Hoover

Senior Editor, InformationWeek Government

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like

More Insights