A number of innovations in the upcoming Windows Vista and Office 2007 products were initially conceived of in Microsoft Research, whose teams then worked closely with the individual product groups to make these technologies viable for inclusion in shipping products.
Among the contributions made by Microsoft Research are the new desktop search, Sidebar and SuperFetch features in Vista, and the new ribbon-based user interface in Office 2007.
One of the new features that has been widely touted by Microsoft executives, including Jim Allchin, the co-president of the Platforms and Services division, and CEO Steve Ballmer, is desktop search, which was conceptualized some five years ago by Susan Dumais, the principal researcher of Microsoft Researchs Adaptive Systems and Interaction group.
This new desktop search technology has a richer UI, quickly searches a users entire computer for queries and keeps the search results in one place.
Dumais, who said she has long been interested in information retrieval, built a desktop search prototype called “Stuff Ive Seen” for Microsoft Research some five years ago. That technology was deployed internally at Microsoft to about 3,000 people, which enabled Dumais and her team to learn about some of the challenges and issues of the project.
“We then began working closely with the MSN and Windows Vista product teams and were able to transfer many of the lessons we learned to them, such as how users know a lot of information about what they are searching for and often remember specific characteristics, as well as the need for a flexible interface and the importance users place on date,” she said.
The new search engine, influenced by the work of Dumais group, is included in the most recent version of MSN and in Windows Vista.
The Windows Sidebar feature is a window that exists on the side of the screen of a Windows Vista PC. It is essentially an application that allows users to display gadgets on the Sidebar itself and on the Windows desktop so they can customize what information appears there, like weather, traffic and stock prices.
The technology started life as a prototype called Sideshow, later renamed Sidebar, which was created by Microsoft Research incubation group program manager JJ Cadiz and his team.
With regard to the SuperFetch memory optimization technology, Eric Horvitz, the principal researcher and research area manager for Microsoft Researchs Adaptive Systems and Interaction group, told eWEEK that one of his personal dreams at Microsoft has been to endow operating systems with intelligence so as to better understand users and context.
Such methods promise to give the operating system the ability to allocate memory and computing resources in a dynamic manner based on predictions of the behaviors and needs of users, he said.
“We worked in tight collaboration with the core operating system team on SuperFetch. We developed algorithms and code that is now part of the SuperFetch component, and we worked closely with passionate people on the operating system team to bring them into the world,” Horvitz said.
These methods give the operating system the ability to observe and learn about the sequences of software applications that users launch over time. By observing users, Vista learns a personalized policy for the prefetching of applications into memory from disk, so as to speed up, on average, the launch of applications.
“That is, when a user clicks to launch an application, the system, which has been learning about the users patterns of activity over time, may often have already paged that application from the hard disk into memory—behind the scenes—so that the user can begin to interact more quickly with the application,” he said.
Machine learning is at the core, with the system “learning” the probabilities of what the user is going to do in the future, by observing sequences of application uses. The system also considers contextual factors like the time of day and the day of the week, Horvitz said.
“We also encoded a model of user frustration with waiting for responses—the system has a sense for growing frustration with waiting—and we use this as part of the triaging of applications that are prefetched into memory so as to be there in advance of the user requesting them. I love the fact that Vista has what we refer to as a preference model: It has encoded, deep within the system, a sense for the frustration that people may feel when they wait for a response,” he said.
The work with SuperFetch is also just the start of efforts to learn about and predict human needs and intelligent resource allocation in operating systems.
“Weve been discussing even more sophisticated methods for future versions of the operating system. Its my hope to see systems continue to use available resources to speculate about what the user might need in the future, and to precompute or prefetch in a very smart way to leverage all available resources all the time,” he said.
While Microsoft Research did not develop the Office ribbon idea per se, Mary Czerwinski, the research area manager of the Human-Centered Computing groups and manager of the Visualization and Interaction Research group, along with her team, helped with some of the brainstorming, analysis and consultation with the program managers, user researchers and developers to provide insight and ideas for the UI.
Very early on in Office 2007 planning, Czerwinski and her team met repeatedly with user researchers and the feature teams, who wanted to reduce the complexity of the features and functions of the Office UI.
“We all eventually decided that some statistical analysis of the way the features and functions of the UI were used might help us. Usage data was analyzed from thousands of volunteers using Office to see what UI features were used most often, and also what features were used in conjunction with others,” she told eWEEK.
These analyses were eventually grouped into task areas by the Office team, and the ribbon was born. After the ribbon was designed, the Microsoft Research team was often consulted on how the ribbon laid itself out for various window sizes, and about how the hot key shortcuts should be designed, Czerwinski said.