A report has set off a ripple across the Internet that Microsoft is set to announce it is buying semantic-search engine provider Powerset. However, so far neither company is commenting on whether that deal is actually happening.

VentureBeat, a news site focus on private companies and their venture capital backers, on Thursday reportedVenturebeat article 26/06 that Microsoft plans to purchase the San Francisco-based startup for a little more than US$100 million.

On Friday, two spokespeople from Powerset said separately that the company does not comment on rumors or speculation. A spokesman for Microsoft's public relations firm also had no comment about any possible deal.

Powerset is pioneering semantic search. Semantic search attempts to extract meaning from search queries and Web pages rather than simply matching them up with relevant links based on keywords or previous or related searches. Google still primarily uses keywords to deliver search results.

Powerset is currently testing a search engine that attempts to understand the meaning of Web pages, in part using technology licensed from Xerox's PARC subsidiary. That technology creates a semantic representation of Web pages by parsing each sentence and extracting its meaning.

If the rumor is true, a deal between Microsoft and Powerset would come as little surprise, as Microsoft needs a way to bolster its search strategy now that its deal to purchase Yahoo has fallen through.

Since search-based advertising is the largest slice in the online advertising pie, Microsoft must increase the profile of its Live Search engine in order to build this part of its business.

Live Search is seen as far inferior to Google's search engine, and also ranks behind Yahoo in terms of how often it's used.

Microsoft Chief Software Architect Ray Ozzie has hinted publicly that Microsoft is looking beyond keyword search to bolster the capabilities of Live Search.

At the Bernstein Strategic Decisions conference in New York last month, he said the company is interested in expanding the ways search engines look for information and how people request searches.

"There is room for improvement with how we structure our requests to find the information we want," he said. "We haven't seen any high-scale implementations for any good, innovative things that have come out."