Powerset Debuts With Search of Wikipedia

Powerset is undoubtedly among the most ballyhooed startups in Silicon Valley. The company’s mission is as improbable as it is daring: just like like Google did a decade ago, Powerset wants to come up with a better way to search.

On Monday, the company will make its public debut, opening up to the masses its search engine, which uses “natural language” to organize and search documents. But Powerset won’t be searching the entire Web, at least not initially. In its debut, it will be confined to Wikipedia.

By using language rather than keywords, Powerset’s demos suggest the company has developed powerful new capabilities. Powerset created an index of Wikipedia by studying the meaning of entire sentences rather than the relationship between words. Similarly, it allows users to type queries as fully formed questions. That allows it to do certain things that many search engines cannot do.

Ask “Who did Henry VIII marry?” or “What did the FDA ban?” or “What did Bill Clinton sign?” and Powerset will come up with remarkably good answers. (Incidentally, Google does a decent job of answering the first of these questions but not the other two.) Powerset also has other nifty features, like its ability to create mini-dossiers that summarize the information it finds and to take users directly to a section of a document that is most relevant to their search.

But Powerset remains a long way off from its promise and faces a seemingly intractable problem: for a very large fraction, if not the vast majority, of searches, keywords work just fine.

“They have a new and interesting technology that most people don’t really need right now,” said Danny Sullivan, a search expert and editor of Search Engine Land. Mr. Sullivan also said that analyzing the meaning of pages, as Powerset does, demands so much computing power that the company is unlikely to be able to index the entire Web any time soon.

Powerset is well aware of that and plans to increase its scope gradually by indexing other “data sets” beyond Wikipedia. News articles and blog posts are likely to be next, the company said.

Given Powerset’s combination of usefulness and limited appeal, it is not surprising that the company is unveiling its product amid rumors that it may be up for sale, with Microsoft mentioned as one possible buyer.

http://bits.blogs.nytimes.com/2008/05/12/powerset-debuts-with-search-of-wikipedia/?hp