DemoFest Booths
-
Presenters: Rane Johnson, Roland Saekow, Katie Doran, Donald Brinkman, Danny Dalal
Learn and try ChronoZoom, an intuitive, online tool for mashing up big multimedia data to gain cross-discipline insight and to bridge the gap between the sciences and the humanities. Explore this master timeline of the cosmos, Earth, life, and human experience, from billions of years ago to today. ChronoZoom unifies a wide variety of data and historical perspectives, enabling researchers, educators, and students to examine historical events, trends, and themes and to synthesize unexpected relationships and historical convergences that help explain the sweep of Big History.
-
Presenters: Sumit Gulwani, Ben Zorn, Dany Rouhana, Shobana Balakrishnan
Microsoft Office is the leading platform for manipulating various forms of data, including strings, numbers, text, tables, and drawings. But identifying the right combination of commands for manipulating such data from Office’s rich set of more than 10,000 features is challenging. Furthermore, several repetitive tasks require programming that is beyond the expertise of most end users. We have mined Office help forums to identify tasks that end users struggle to accomplish, and observed that users often expressed their intent by using examples. We have developed a program-synthesis technology for automating repetitive data-manipulation tasks from examples.
We will demonstrate add-ins for Excel and Word that enable automation of a variety of tasks from examples such as converting dates in multiple formats into a desired format, semantic search-and-replace in text documents, and layout and formatting transformations on tables. We also will show a prototype tool for PowerPoint and Visio that enables the beautification and prediction of structured or repetitive geometric drawings.
-
Presenter: Joseph Joy
Rich Interactive Narratives (RIN) is a presentation technology for creating, representing, and presenting experiences that merge cinematic narrative with interactive exploration over an extensible range of rich media. The RIN project aims to combine the compelling, time-tested narrative elements of multimedia storytelling with the information-rich and exploratory nature of the latest generation of information-visualization and -exploration technologies. We approach the problem not as a one-off application, Internet site, or proprietary framework, but rather as a new kind of semantic representation, one that transcends a particular platform or technology. We will showcase the technology through demonstration of the authoring experience, as well as content created with collaborators. Specially featured is an interactive narrative on the plight of the mighty Himalayan glaciers, built in collaboration with accomplished mountaineer and award-winning cinematographer David Breashears and the nonprofit organization he founded, GlacierWorks.
-
Presenters: Alex Wade, Greg Tananbaum
Microsoft Academic Search is a free service that was developed by Microsoft Research to help scholars, scientists, students, and practitioners quickly and easily find academic content, researchers, institutions, and activities. Microsoft Academic Search indexes not only millions of academic papers, but it also surfaces key relationships between and among subjects, content, and authors in a manner that highlights the critical links that help define scientific research. Microsoft Academic Search also serves as a test bed for many research ideas in areas such as data mining, named-entity extraction and disambiguation, and data visualization.
-
Presenter: Kristin Tolle
The purpose of the DataUp system, developed in collaboration with the California Digital Libraries and the DataONE users group, is to enable Microsoft Excel spreadsheets on local computers to be uploaded seamlessly to repositories along with associated metadata and a unique data identifier for later access and updating. This ensures data preservation of the myriad of spreadsheets spread across many users and collaborators, the capture of repository complaint metadata, and support for data management required by funding agencies. There is both a web service and an Excel add-in version of the application that is targeted for initial use by the environmental-sciences community. These open-source applications will be made available by the Outercurve Foundation.
-
Presenters: Dean Guo, Kristin Tolle
Microsoft Translator Hub empowers businesses and communities to build, improve, and deploy customized, automatic language-translation systems—bringing better, specialized translation quality to established languages, as well as the many native languages of the world that are not yet supported by major translation providers. Powered by Windows Azure, Microsoft Translator Hub is an extension of the Microsoft Translator platform and service. You can build a superior translation system easily within a private website by combining your translated documents with the power of Microsoft Translator’s big-data back end. Once you are satisfied with your translation, you can deploy the system and share it publically on the web using the webpage widget and/or access it privately through the Microsoft Translator APIs.
-
Presenters: Danyel Fisher, Rob DeLine, Steven Drucker
Big-data analytics require new workflows; high-latency queries, massively parallel code, and cloud-computing infrastructures make handling a big data set different than working on a local machine. Exploratory data analysis can become impossible when queries take a long time. We are exploring user experiences for analysts and thinking about new ways to cope with big data sets.
We will demonstrate a tool that enables users to visualize the results of progressive queries of large data sets and an interactive scripting environment for developing progressive queries against large data sets. These tools enable users to test multiple hypotheses and explore their data.
-
Presenter: Patrick Longa
Many cryptographic systems—most notably, public-key cryptosystems—are known to be computing-intensive and high-power consumers. To fulfill the increasing demands of privacy and security for a wide range of applications such as cloud computing, efficient realization of these cryptographic systems is critical.
We have developed novel techniques and applied algorithmic optimizations to speed the computation of most expensive elliptic-curve and pairing-based cryptographic operations, and then implemented them efficiently on x64 platforms.
First, we illustrate, with an example based on batch signatures, how our techniques can significantly speed sample server workloads when performing cryptographic signature verifications. Next, we show the performance improvement that is possible when computing a batch of pairings, as in potential applications for cloud services.
-
Presenter: Shelly Farnham
So.cl is an experimental social network from FUSE Labs that reimagines search as social from the ground up. With So.cl, you can share your interests by posting images and links you find on the Internet into rich visual posts. Users find new interests and connect with like-minded people around the world. So.cl was released in a private beta in December 2011, and we have been iterating the design and nurturing the community ever since.
-
Presenter: Matej Ciesko
The understanding of parallel programming and developing applications that efficiently exploit the underlying parallel architectures have become requirements for mainstream software engineering. As a result, parallel programming has moved into the focus of university curricula. The National Science Foundation (NSF) is supporting the nationwide distribution of University of California, Berkeley, course materials that are related to parallel programming.
The NSF Parallel Programming Course aspires to contribute to this NSF/Berkeley program by providing an alternative for faculty and students wishing to learn parallel programming on Windows and Microsoft Visual Studio. University departments adapting this material for courses that are offered on parallel-programming technologies will benefit from a richer assignment portfolio covering traditional approaches to parallel computing, as well as today’s new/emerging technologies, such as accelerators.
This demo illustrates how students can use course materials and the Windows platform to acquire an understanding of parallel programming.
-
Presenter: Derrick Coetzee
Domain-expert productivity programmers desire scalable application performance, but to achieve it, they must rely on efficiency programmers who are experts in explicit parallel programming. Because such efficiency programmers are rare, we describe a methodology known as Selective Embedded Just-in-Time Specialization (SEJITS) that encapsulates their strategies in mini-compilers for domain-specific embedded languages (DSELs) hosted in a common, high-level host language familiar to productivity programmers. The mini-compilers emit source code in an efficiency language that targets low-level hardware optimizations using downstream compilers and auto-tuning. The result is source and performance portability for productivity programmers—and performance that rivals that of hand-coded efficiency-language implementations of the same applications on each platform.
We will show a Python image-processing application that uses SEJITS’ stencil DSEL targeting C++ with OpenMP. It achieves performance better than hand-tuned C++ and within a few percent of the best achievable performance based on memory-bandwidth limits.
-
Presenter: Lucas Bordeaux
We have developed a tool that “synthesizes” code snippets from natural language queries. The queries usually are related to API usage, restricted for now to C#, for instance, “How do I generate a SHA1 hash?” or “How do I read a file line by line?” From these questions, we compute a code snippet that solves the question and exemplifies best practices. This piece of code is a simplified “code skeleton” that the user can fill in easily.
-
Presenter: Rob Fatland
We will demonstrate new visualizations—home-grown and contributed—from the Layerscape data-visualization project powered by WorldWide Telescope. In the context of our environmental-science “data-plus-tools” vision, we will showcase “how we got there” by using resources from Microsoft Research Cambridge, Microsoft Academic Search, and the Microsoft technology stack. We also will present “where we go next,” specifically, ideas on the future of location-based predictive modeling.
-
Presenter: Ece Kamar
Crowdsourcing offers opportunities for harnessing human intelligence to solve tasks that computers cannot do alone. The manual work needed for ensuring quality for crowdsourcing systems is a major limitation for large-scale applications. Systems that can automate crowdsourcing and provide quality guarantees could enable the widespread use of crowdsourcing and change the way we collect data, the way we search the web, and the way we perform tasks.
We introduce CrowdSynth, an end-to-end system for automatically solving and managing crowdsourcing tasks. CrowdSynth uses machine learning and inference to take advantage of the complementary strengths of people and computational agents to solve crowdsourcing tasks. The system assesses the quality of a task continuously and makes hiring decisions by combining techniques from machine learning and decision-theoretic optimization. We tested the system on a large-scale citizen-science project and showed that the system can achieve significant savings in the resources needed to solve crowdsourcing tasks accurately.
-
Presenters: Michael Zyskowski, Bob Walter, Evelyne Viegas, Kenji Takeda
TryF# is currently undergoing development for the next big iteration of web-based programming for learning the F# language. This version is meant to showcase the power of F# to solve real-world analytical programming and information-rich problems by taking advantage of the power of F# 3.0 Type Providers, as well as providing more advanced web-based code editing features, more complex program capabilities, and cross-platform support. This demo will provide an early view of work in progress and the opportunity for community feedback.
-
Presenters: Scarlet Schwiderski-Grosche, James Scott, Kerry Hammil, Nicolas Villar
Microsoft .NET Gadgeteer is an open-source toolkit for building small electronic devices by using the .NET Micro Framework and Visual C# Express, the free C# development environment. .NET Gadgeteer combines the advantages of object-oriented programming, solderless assembly of electronics with a kit of peripherals, and support for quick form-factor construction using computer-aided design. This powerful combination enables embedded and handheld devices to be iteratively designed, built, and programmed in a matter of hours rather than days or weeks. Microsoft .NET Gadgeteer has been developed by Microsoft Research in collaboration with the .NET Micro Framework product team. Commercial availability of .NET Gadgeteer was announced in August 2011, and hardware is available from a number of hardware manufacturers. Since autumn 2011, .NET Gadgeteer has been used in schools and universities for teaching and research.
We will provide an overview of the various hardware and software elements of the.NET Gadgeteer platform. Attendees will be introduced to the modular electronics system and learn how individual modules can be connected and programmed easily by using the C# language—yielding sophisticated devices. We also will demonstrate a range of devices built with .NET Gadgeteer as part of research collaborations and teaching engagements with academia.
-
Presenters: Ratul Mahajan, AJ Brush, Rayman Preet Singh
HomeOS advances the abstraction of a home infrastructure as one computer rather than a collection of devices. It simplifies how users grow and manage their home infrastructure and how developers write applications for devices in the home. We and our academic partners will demonstrate systems and applications for the home that were built by using HomeOS.
-
Presenters: Nikolai Tillmann, Peli de Halleux, Manuel Fahndrich
TouchDevelop is a new, mobile development environment that enables anyone with a Windows Phone to create apps directly on the smartphone. At the core is a new mobile programming language and editor designed with the touchscreen as the only input device. TouchDevelop embraces an ongoing technology shift: instead of PCs and laptops, mobile devices are becoming more prevalent for most everyday computing tasks. In fact, never before in human history have incredibly powerful, versatile computing devices such as smartphones been available and adopted so broadly. Programs written in TouchDevelop can take advantage of the computing power of the smartphone and all its sensors, such as GPS, camera, accelerometer, and gyroscope, and stored personal data, such as contacts, songs, and pictures. Thousands of users already have written TouchDevelop programs for fun. For students, programming on mobile devices promises to be an engaging experience.
-
Presenters: Ben Glocker, Darko Zikic
We will demonstrate two clinically important applications solved via a powerful machine-learning technique called decision forests.
The first application addresses the automatic delineation of brain tumors in multichannel magnetic-resonance images. The developed algorithm enables quantification of tumors, which is important for assessing disease progression and success of treatment.
The second application aims for automatic detection and localization of individual vertebrae in arbitrary field-of-view computed-tomography scans. Spinal anatomy serves as an important, patient-specific coordinate frame that is determined automatically by the proposed system. This enables many subsequent analysis tasks, such as guided visualization, shape and population analysis, and anomaly detection.
These applications demonstrate the general purpose and wide applicability of decision forests—and the potential of machine learning for medical-image analysis.
-
Presenter: Hrvoje Benko
Wearable Multitouch Interaction is a depth-sensing and projection system that enables interactive multitouch applications on everyday surfaces. Beyond a shoulder-worn system, there is no instrumentation of the user or the environment. Foremost, on such surfaces—without calibration—Wearable Multitouch Interaction provides capabilities similar to those of a mouse or a touchscreen: X and Y locations in 2-D interfaces and whether fingers are “clicked” or hovering, enabling a wide variety of interactions. Reliable operation on the hands, for example, requires buttons to be 2.3 centimeters in diameter. Thus, it is now conceivable that anything one can do on today’s mobile devices can be done in the palm of a hand.
-
Presenter: Andy Wilson
Holoflector is a unique, interactive, augmented-reality mirror. Graphics are superimposed correctly on your reflection to enable an augmented-reality experience unlike anything you have seen. We also use the combined abilities of Microsoft Kinect and Windows Phone to infer the position of your phone and to render graphics that seem to hover above it.
-
See innovative student design solutions from the Carnegie Mellon University School of Design, Umeå University in Sweden, London’s Royal College of Art, the Delft University of Technology in the Netherlands, Tsinghua University’s Academy of Arts & Design in Beijing, Escola Superior de Desenho Industrial in Rio de Janeiro, and the University of Washington. The student design solutions include:
- Papercake, a comprehensive social-sharing system that integrates financial, medical, career, and social information—from Carnegie Mellon University.
- Lango, technology that enables new speakers of a language to learn, understand, and progress through its cultural meanings—from Umeå University.
- Two projects that use algorithms to find patterns and cultural meaning in everyday events—from the Royal College of Art.
- Elevator, a project that helps people engage in social-sharing experiences in a universally awkward social situation: the long elevator ride—from Delft University of Technology.
- Roommate, a system intended to make it easier for graduates to adapt to post-school life by helping them find affordable housing—from Tsinghua University.
- Zome, a system and visual language that helps people access and understand personal recommendations—from Escola Superior de Desenho Industrial.
- APT, a service that delivers apps that are relevant to an individual’s location and preferences—from the University of Washington.
-
Presenters: Ranveer Chandra, Hagen Green
The wireless display is a rapidly emerging technology in new-generation portable devices. It provides the ability to stream the contents of a screen wirelessly to a remote device display, thereby enabling interesting scenarios such as video streaming, gaming, and screen mirroring. Even as we see initial solutions coming to market, there are a couple of key concerns that don’t seem to be addressed. First, current solutions do not consider relatively low-power mobile devices, such as mobile phones, as the Wi-Fi display source. Simply porting the technology to such devices can create serious energy concerns. Second, wireless displays have the potential to unearth interesting use cases for application developers. Current solutions, though, do not appear to provide simple hooks to developers to enable their use cases.
We explore techniques on how to make wireless displays more energy-efficient. We examine interesting scenarios in this regard and propose intelligent buffering, along with other techniques, to mitigate some energy concerns. We also study the current media pipeline (DirectShow) in Windows Phone and suggest how wireless displays can be enabled in the current framework. Further, we take a more orthogonal approach to this problem by wearing the application developer’s hat and exploring how some of these scenarios can be enabled at the Silverlight layer.
-
Presenter: Larry Heck
Project Louise is an ongoing, multi-year prototyping effort within the Microsoft Conversational Systems Labs to build an automated personal assistant. Previous efforts have yielded a multi-domain personal assistant on Windows Phone incorporating a calendar, reminders, restaurants, movies, and events. We will show Project Louise extended across multiple screens, including television. A conversational-interaction model, along with support for multimodal input such as speech plus gesture, has been developed to provide a novel conversational, natural-user-interface experience. The system operates seamlessly over the breadth of the Internet, with a combination of web browsing, search, and deep, rich experiences in a handful of high-value domains, such as those developed previously for Windows Phone.
-
Presenter: Brian Meyers
The CloudTop prototype explores the user experience of linking all your personal computers into a “personal cloud.” The user can drag running programs from one machine to another. Programs, such as Microsoft Word or Excel, are available on demand, rather than requiring the user to install them to each machine in advance. If your laptop gets overloaded, CloudTop automatically can load balance with your desktop. If a machine crashes or runs out of battery power, CloudTop automatically migrates your work to another machine, and you can resume right where you left off.
DemoFest Map
