Projects


Supporting Complex Search Tasks Through Suggestion & Serendipity

PRINCIPAL INVESTIGATOR: Charles Clarke, CO-INVESTIGATORS: Gordon Cormack

ABSTRACT: Traditional search engines, such as Google and Bing, usually provide excellent results when their users have clearly defined and straightforward requirements ("I want to buy a plane ticket" or "I want to learn about the University of Waterloo"). Unfortunately, users with complex and vaguely defined requirements may not be as successful. These users may have to slice-and-dice their needs into several queries, laboriously combining the answers to dig for a solution. This problem may be particularly true for users searching news, video, and other digital media considered under this proposal, as well as for users of mobile devices, where the screen is small and time is short. Imagine a user with a complex and vague query, such as "entertain me". If the user is a tourist in a city they don't know, a search application might help them plan their evening, locating suitable restaurants/theaters/clubs, looking at reviews, locations and distances, individual preferences, and timetables. If the user is sitting at home, the search application may suggest videos, games, news, or even a night on the town. Keys to solving this problem include the ability to interpret behavior, integrate context, suggest possibilities, and summarize content. Elements of this proposal address each of these aspects, all within the broad goal of supporting complex search.

Low-Level Low-Overhead Program Tracing for Multimedia Application

PRINCIPAL INVESTIGATOR: Sebastian Fischmeister, ECE CO-INVESTIGATORS: Hiren Patel, ECE

ABSTRACT: Multimedia applications require different features than standard desktop software. For example, while users may accept short wait periods on desktop software, multimedia applications like a video player are time-sensitive where a delay in displaying a frame results in poor video experience. Furthermore, the replay must occur without dropping many frames. Building such time-sensitive applications is notoriously difficult. For example, correct replay depends on the correct display of frames as well as the timeliness of the application. Showing correct frames too quickly or too slowly in sequence still results in an unpleasant viewing experience. To help building such applications, developer often use program tracing. Program tracing is a well-known technique for ensuring that applications follow the specification. Program tracing observes and logs all actions of the application for later inspection. Tracing naturally affects the performance of the application, and in the worst-case, tracing can result in software defects occurring only when the system trace is disabled. Consequently, it is important to investigate program-tracing mechanisms that exhibit low overhead at runtime, but can still operate at a low level in the system. Since next-generation multimedia devices will be a mix of special hardware-accelerated components and software applications, we will investigate in our work the hardware as well as the software side of the problem of program tracing.

Privacy-friendly Context-aware Recommender System

PRINCIPAL INVESTIGATOR: Urs Hengartner

ABSTRACT: Imagine it is Friday late afternoon and you are wondering how to spend the evening. You grab your smartphone, which tells you that there is a jazz concert taking place not far away from you and that several of your friends are already there. Since you are a Jazz aficionado, you are more than happy to follow your phone's suggestion. You ask your phone for directions to the venue and you get to spend a nice evening with your friends. The described application is an example of a recommender system that takes advantage of a user's context (location and time in the example), the user's interests (jazz), and the user's social network (friends). Given that this kind of information is often available to an online social networking site, such as Facebook, there is little doubt that such recommender systems can and will be built. However, privacy-conscious users would be hesitant to reveal all this potentially sensitive information to a single site and therefore could not benefit from such a recommender system. The goal of the proposed research is to design and build a privacy-friendly context-aware and social network-aware recommender system.

Automation of aesthetic conventions in comics, film and games

PRINCIPAL INVESTIGATOR: Craig S. Kaplan, CO-INVESTIGATORS: Avrim Katzman, John Helliker

ABSTRACT: Film, games and comics are all artforms that are governed by aesthetic conventions. These conventions have evolved organically over long periods, and typically repre-sent hard-won knowledge of how to communicate effectively with an audience. We will explore how to enhance the software systems typically used in film and comic production in order to encode, communicate, and enforce the aesthetic conventions. The ability to work within such conventions (or to defy them when desired) will al-low a content creator to articulate their vision more precisely and more expressively, to avoid common pitfalls, and to spend more time on the high-level creative tasks of their chosen medium.

Personal-Public Display Interaction and Integration

PRINCIPAL INVESTIGATOR: Edward Lank

ABSTRACT: Modern slate/tablet displays and devices all support two-dimensional interactions, typically via multi-touch input hardware. However, these bibliographic and epigraphic multi-touch surfaces seem to focus more on accessing information, i.e. on information output, than on content creation. Even typing is awkward on these devices given the lack of tactile feedback associated with keypress events and the sensitivity of soft keyboards to accidental brushes with fingertips. The goal of this sub-project is to explore content creation on multi-touch displays. We focus on the gesture as the primary vehicle for content creation and the sketch, doodle, or drawing as an initial example of a graphical artifact that can be easily constructed by end-users of multi-touch interfaces. In this sub-project, we explore three specific research questions: - What is the role of sketching, drawing, and doodling on both large public and small personal slate displays? - On shared public displays, how can we best design systems that allow personal interaction within the public sphere? Should we appropriate space on public displays for small groups? If so, how do we negotiate space constraints between one or more groups and the shared public nature of content? - Given the multi-device nature of the world, what styles of interaction might be designed that take advantage of personal devices and public devices working in conjunction? How might these interactions change given various device affordances?

Interactive Narrative on Mobile Devices

PRINCIPAL INVESTIGATOR: Michael Terry

ABSTRACT: Modern mobile devices are specifically designed and tuned for media consumption, as opposed to general computation. However, unlike the traditional media they seek to replace (e.g., books), mobile devices can deliver dynamic, interactive content.This research will develop the medium of interactive narratives on mobile devices, with a particular focus on educational narratives. Just as modes of storytelling evolved to exploit the unique affordances of film, radio, and TV, storytelling on mobile, interactive devices must also evolve to match the affordances of this new technology. This research will examine how the rich sensor inputs (e.g., touch, location, orientation, sound, video, and audio inputs) and network capabilities of modern mobile devices can be leveraged to create new interactive narrative experiences on these devices.


Campaign Waterloo

Digital Media Enabling Technologies Project
c/o Institute for Computer Research (DC 1330)
University of Waterloo
Waterloo, Ontario, Canada N2L 3G1

Tel: 519-888-4530
Fax: 519-746-5036

Contact | Feedback: tamer.ozsu@uwaterloo.ca | Digital Media Enabling Technologies


Valid HTML 4.01!Valid CSS! Last modified: Friday, 26-Aug-2011 13:51:19 EDT


Menu:ShowHide