|
|
 |
Home > Research > Themes > Models of Information in Use
Models of Information in Use
Researchers in this area focus on formal treatments of information processes such as information theory, probabilistic models and stochastic processes, complex systems, and artificial intelligence methods, especially as related to real-life information problems.
Researchers
Current Projects
Past Projects
Agent-Based Models of Cooperation
This ongoing series of computer simulation studies investigates structural and communication conditions that make it more likely that model actors will cooperate in various dilemma games. Collaborators in this project are Rick RIolo of the Center for the Study of Complex Systems, Robert Axelrod of the Department of Political Science and School of Public Policy, and Michael Cohen of the School of Information and the School of Public Policy. The project has yielded publications in such journals as Nature, Rationality and Society, and the Journal of Economic Behavior and Organization.
Contact: Michael Cohen (mdc@umich.edu)
|
Augmenting Expertise Networks
Researchers are studying expertise networks -- the technical augmentation of the ad-hoc social networks by which people seek information, answer questions, and accomplish tasks. Expertise networks could allow organizations to better share information and skills and to allow people to more easily find the expertise they need. Expertise networks could also provide digital libraries and scientific collaboratories with the much-needed capability to obtain informal information. Within these networks, moreover, one can have the flexibility to include information databases, documents, agents, and people together as resources.
Contact: Mark Ackerman (ackerm@umich.edu)
|
Bridging the Gap: End-to-End Networking for Landmark Applications
This project involves the planning of a two-day workshop to address the "Wizard Gap" present in networking problems that the scientific application community experiences It will be the first workshop to simultaneously address this gap from both the network engineering standpoint and the network expert (or "wizard") point of view. Its significance is in bringing together these experts with members of scientific research communities that use demanding, network-driven applications and with developers of these applications and campus network engineers. Researchers hope to better understand the coordination and requirements necessary to support complex and collaborative end-to-end performance troubleshooting and to advance research into the development of tools to support this troubleshooting.
Contact: Barbara Mirel (bmirel@umich.edu)
|
Complex Systems
Several School of Information faculty work closely with the group in
the Center for the Study of Complex Systems on projects ranging from organizational learning, adaptive systems, and political economy to computational markets and Web models.
Contact: Michael Cohen (mdc@umich.edu)
|
Computational Linguistics and Information Retrieval
(CLAIR)
The research focuses on work in two main areas: developing question answering systems that can automatically find answers within a vast amount of underlying text, and developing text summarization systems that automatically identify salient concepts in a text narrative, conceptualize the relationships that exist among those concepts, and generate a concise representation or summary of the input text that preserves the gist of the original document (or documents). An example of a project that has come out of this research group is NewsInEssence, a free, Web-based service that automatically collects and summarizes multiple, related news stories about any topic or event you choose.
Contact: Dragomir Radev (radev@umich.edu)
|
Creative Archiving at Michigan and Leeds: Emulating the Old on the New
(CAMiLEON)
A team of researchers at the University of Michigan and research staff in the United Kingdom from the Cedars project, being run at the universities of Leeds, Oxford, and Cambridge under the aegis of the Consortium of University Research Libraries, won funding from the National Science Foundation in the U.S. and the Joint Information Systems Committee in the UK for an international digital library intiative to investigate the potential role of emulation in long-term preservation of digital resources. The project developed a small suite of emulation tools, evaluated the costs and benefits of emulation as a preservation strategy for complex multi-media documents and objects, and developed models for collection management decisions to assist people in making real life decisions about how much effort and resources to invest in exact replication within preservation activity (as opposed to preserving raw intellectual content). The team developed preliminary guidelines for the use of different strategies (conversion, migration, and emulation) for managing and preserving digital collections.
Contact: Margaret Hedstrom (hedstrom@umich.edu)
|
Designing Online Interactions to Foster Interactive Analysis and Decision Making
Contact: Barbara Mirel (bmirel@umich.edu)
|
Models for Natural Language Processing
The Computational Linguistics and Information Retrieval group works on statistical graph-based models for semi-supervised learning (e.g., classification with a very small number of labeled examples).
Contact: Dragomir Radev (radev@umich.edu)
|
Models of the Web
Researchers are interested in building adequate generative models of the Web that explain its structure (in terms of the degree distribution of pages, the characteristic path length between pages, and the clustering coefficient of the Web graph). More specifically, they have tried to account for the role of lexical information (the presence of particular words on pages) in deciding the local
connectivity of the Web graph.
Contact: Dragomir Radev (radev@umich.edu)
|
Multi-Agent Systems
Researchers in this area work to develop intelligent coordination among multiple (semi)autonomous systems involving the proactive selection (planning) of physical, communicative, and/or computational actions that improve performance in a multi-agent context. The work thus is concerned with how artificial agents should decide what courses of action to commit to given a multi-agent world, how they should meet those commitments (including meeting real-time constraints), and how they should revise and renegotiate their commitments based on unexpected events in their environment.
Contact: Edmund Durfee (durfee@umich.edu)
|
Multi-Mode Image Retrieval Group
Research focused on a way to develop and evaluate image retrieval techniques that use both visual and textual cues to query image databases. Such content-based image retrieval methods would enable users to search for images based on image features, such as color, shape, and texture. One of the group's primary goals was to develop retrieval techniques that focused on the needs of generalist or naive users. The testbed for the group's work was The Earth and Space Science Browser, a Web- accessible digital image library containing more than 1,400 images of the earth, planets, star systems, space craft and other earth and space science subjects.
Contact: C. Olivia Frost (cfrost@umich.edu)
|
University of Michigan Digital Library Project
(UMDL)
The project incorporated basic research, system building, deployment, and evaluation of a usable digital library. The UMDL provided access to significant public domain collections of scientific data and information. The University received funding for the project from the National Science Foundation, the Defense Advanced Research Projects Agency, and the National Aeronautics and Space Administration.
Contact: JoAnne Kerr (jmkerr@umich.edu)
|
Home > Research > Themes > Models of Information in Use
|
|
|
|
|
|
|
Associate Professor Margaret Hedstrom is prominent nationally and internationally in efforts to ensure that the torrent of "born digital" documents now being created -- including what are in effect the birth records of the information age -- are preserved and archived in ways that will make them accessible for future generations.
|
|
|
|
|
|
