Microsoft researchers dream big

MOUNTAIN VIEW, Calif.--What if your computer could read the newspaper for you and tell you just what you needed to know? What about 1,000 newspapers?

Researchers at Microsoft think computers can do just that. Someday.

Google and Microsoft already offer news aggregation sites that collect all of the day's news stories from throughout the Web. However, MSN Newsbot and Google News present the first part of a story, assuming that is the most important part.

Microsoft research is looking into creating software that can read the whole article--and dozens of others on the same topic--and come up with an accurate summary. In theory the research could be extended to even allow the summary to mimic a particular style--giving, say, a Wall Street Journal-style summary of Jennifer Lopez' wedding to Marc Anthony.

"The research issue is comprehension," said Lucy Vanderwende, who is heading the project. "Can the machine understand?"

If the answer is yes, you can bet the folks at MSN want to know. Indeed, several of the research projects Microsoft showed off Wednesday in Silicon Valley are designed to help the company in two of its most critical challenges--battling security threats and taking on Google in the Web search and information business. The miniature science fair was part of a road show for Microsoft, which held a larger exhibition in March at its main campus in Redmond, Wash.

Whether it is summarizing news stories or using statistics to find out which Web pages are out there to fool search engines, MSN seems to be the most likely beneficiary of much of the Web-related research efforts.

"Their interest is pretty high," Vanderwende said, although she noted that the need for the document summary technology she is researching extends well beyond news into academics and government as well.

In the area of security, the basic research can't come fast enough, admits Michael Schroeder, the assistant director of Microsoft's Silicon Valley research unit, which is based at Microsoft's campus here.

"In some cases, the product group is waiting for the code drop every Friday," Schroeder said. "There are a set of problems around Trustworthy Computing that are just crucial."

One of those areas is a highly touted concept called shield technology. Microsoft's highest executives have promised that the idea will find itself soon into the core Windows operating system. The idea is that when there is a vulnerability, a shield can provide a Band-Aid that makes sure nothing gets in the wound until a full patch can be applied to close the hole.

Much of the fundamental research in the area, however, is still ongoing, even at Microsoft. Helen Wang is part of a team of researchers trying to advance the state of the art, even as the Windows teams scrambles to include the technology in forthcoming versions of the operating system.

While shields will not prevent exploits that occur just as a vulnerability is discovered, Wang notes that 90 percent of attacks exploit a well-known vulnerability. In most cases a patch is available before the attack, but may not be widely installed.

"People do not patch timely," Wang said.

Rather than fight a losing battle trying to convince companies to change their ways, Wang said shields can offer protection without the same need for testing. In some areas, shields could replace patching altogether. In the pharmaceutical industry, for example, companies have to disclose to regulators any patches they do, but the same might not be true of shields.

It's that kind of "street smarts" that seems characteristic of Microsoft's researchers. To thrive inside the software maker's lab, one also has to be ready to shift one's field of study.

"The set of things that are worth (looking into) shifts as time goes by," Schroeder said. "A good researcher has to be cognizant of that and be flexible."

However, even with the attention being given to near-term problems, Schroeder says Microsoft's research unit isn't going to lose its long-term focus. He said there's a healthy balance of work that is 10 or 15 years from having a practical application, work that is useful immediately, and all sorts of projects in between.

"Any good lab is going to have a mix," he said.