Microsoft buys semantic search engine, Powerset
By TJ Kirchner
Microsoft recently purchased the Powerset, a search engine company based in Silicon Valley that can processes natural-language questions and produces a list of answers based on Wikipedia articles. After their recent attempts to acquire Yahoo’s search engine, perhaps it’s trying to use this technology to upgrade Live.com
The price of the acquisition is rumored to be around $100 million. An official announcement will be released next month. Microsoft’s Steve Ballmer had this to say.
The most important application for the foreseeable future is search… We don’t have to dominate, but we’d better have a darn good chunk of the search market over time, and we’re working away at it.
iPhone users will be happy to know that the Powerset application is already available for them to download. So far, feedback has been positive. Customers say that this app. is better than its web version.
Granted the idea behind this technology is nothing new. Ask Jeeves (before it became Ask.com) had the same ability of processing semantics and producing a list of search results of where the user could find their answer.
However, most people were annoyed when the search engine listed irrelevant sites. Since then, Ask.com focused on keywords, much like what Google does.
Google’s approach to semantic searches has usually been rather limited. If the user asks a question that produces a numeric answer or uses an asterisk for a fill-in-the-blank search, the engine may produce the correct result. Powerset takes this a step further by attempting to understand the full meanings of phrases and referencing the largest online encyclopedia, Wikipedia.
Unfortunately, the company has faced some difficulties in the past. Pulling off semantic searches is a lot harder to do than it sounds. For that reason, skeptics within the industry doubted if the technology could be created and used in a major search engine. Plus, it’s been very costly and time consuming to develop.
Furthermore, just like Ask Jeeves, it doesn’t always answer the questions correctly. It’s a lot more accurate in answering “who” or “what” questions rather then “when.”
Given the fact that it’s drawing information from Wikipedia, this makes some sense, since the bulk data is based on people, places, and things and not so much on individual years. Plus, there’s the questionable accuracy of Wikipedia’s content to consider.
All and all, I think this is a step in the right direction for Microsoft to differentiate them in the search engine market. Normal users are more likely to type questions or statements into a search bar rather than a couple keywords and see what comes up.
Related:





Stumble It!

June 30th, 2008
Its important that search applications emphasize the myriad subtleties in information classification to cater to different user characteristics. When these differences power search engines, we’ll get closer to perfect representation. Take a look at http://www.topodia.com = interesting approach