Microsoft touts research projects in search, privacy

Characterizing research as critical to the company’s survival, Microsoft officials Thursday offered glimpses into multiple research efforts in areas such as privacy and search.

In an event at Microsoft Research’s Silicon Valley facilities in Mountain View, Calif., Microsoft officials said the company has 2,500 people doing research and development in the valley. Globally, the company has more than 800 PhD holders involved in research.

“Ultimately, the goal of Microsoft Research is to make sure there’s still Microsoft in 10 years. We’ve been able to do that for 17 years,” since the beginning of Microsoft’s research enterprise, said Rick Rashid, senior vice president of Microsoft Research.

Microsoft was a small company, with only about $1 billion in sales, when it decided to build a research organization, Rashid said. The Silicon Valley facility focuses on distributed computing.

Research, Rashid said, gives Microsoft a chance to respond to a changing competitive environment. Basic research also offers society and humanity the opportunity to survive, he said.

Tongue in check, he stressed the advantages of having a research group should something really bad happen, such as a war, famine, or Google, which is located just a few minutes’ drive from Microsoft’s Mountain View offices.

Microsoft has tried to acquire Google competitor Yahoo. In an interview, Rashid said Microsoft has not been competing with Google for researchers lately because Google does not do any basic research.

Indeed, Microsoft Research efforts in search — Google’s forte — were evident at Thursday’s event. With its query-dependent ranking project, for example, Microsoft has been evaluating algorithms to get improved relevance of Web search results. The Keyboard Generation and Query Classification project at Microsoft focuses on developing technology to show keywords to advertisers.

In the privacy arena, the company’s Privacy Integrated Queries project is intended to enable queries on data while protecting certain information, such as a person’s health history. It leverages Microsoft’s Language Integrated Query (LINQ) and Differential Privacy technologies.

“Differential Privacy we believe is a very powerful idea for building systems that guarantee privacy in a provable, mathematic sense,” said Roy Levin, Microsoft Distinguished engineer and director of Microsoft Research – Silicon Valley.

“There are benefits to sharing data, but you may not want it revealed,” Levin said.

With Privacy Integrated Queries, information such as a particular employee’s salary could be obscured from a person making several queries in a deductive manner in order to find out this private information, Levin explained.

As part of the company’s electronic market design focus, efforts are under way to design algorithms for use in applications such as auctions for digital goods and time consideration in auctions. Also being researched are shortest-path algorithms to find optimal travel routes.

Programmer improvements were a focus as well. One project cited, DryadLINQ, is intended to make it simpler to do large-scale data parallel computing. A DryadLINQ program is a sequential program featuring declarative LINQ queries executing arbitrary sequential code. A program is translated into a distributed execution plan, with easy programming of data mining and other data-parallel computations, Microsoft said.

To assist programmers programming for multiple-CPU systems, Microsoft’s Automatic Mutual Exclusion project automates operations programmers have had to do manually via the mutual exclusion method.

The Microsoft MoDist effort, meanwhile, features a tool to identify bugs in distributed systems. The project involves “in-situ” model checking and intercepting calls through WinAPI on each machine. Also in the distributed systems arena, Microsoft’s Constellation project seeks to automate solutions to failed operations in a networked environment.

Microsoft is researching system architecture to find a solution to the “Boolean satisfiability,” or SAT, issue. “This is another classic problem in computer science,” Levin said. This effort would involve solving large sets of equations to offer an advantage in applications such as financial markets.

A project entitled, “Bilingual Built-ins That Break Language Barriers” features applications of translation technology that shows how machine translation integrated in Microsoft products can help eliminate barriers to worldwide communication. With this online service now in an incubation stage at Translator.live.com, cross-language hurdles are addressed.

As part of this effort, a multilingual chat service is being developed, to be launched in a couple of months. It is available in an early development phase for Windows Live Messenger users at this e-mail address: mtbot-en_us.hotmail.com.

With the BEE3 (Berkeley Emulation Engine) project, Microsoft is researching FPGAs (Field Programmable Gate Arrays) as a way to advance computer architecture research.

Also highlighted were WorldWide Telescope, combining imagery from space-based telescopes with Internet data, and E-Science in the Cloud, to boost eco-science efforts.

(Robert McMillan of IDG News Service, an InfoWorld affiliate, contributed to this report.)