Digital Unwrapping of Ancient Scrolls
Researcher: Dr. Brent Seales
From ancient times, words have been recorded that express immortal ideas and thoughts about society, culture and philosophy. Around the world, people have recorded these writings in a variety of ways. However, we are now more aware than ever of the possibility of losing these recordings of human history.
Seales and his team are using 21st century technology to preserve the traces of ancient cultures before the relics disappear forever. The EDUCE project (Enhanced Digital Unwrapping for Conservation and Exploration) is developing a hardware and software system for the virtual unwrapping and visualization of ancient texts. The overall purpose is to capture in digital form fragile 3D texts, such as ancient papyrus and scrolls of other materials using a custom built, portable, multi-power CT scanning device and then to virtually "unroll" the scroll using image algorithms, rendering a digital facsimile that exposes and makes legible inscriptions and other markings on the artifact, all in a non-invasive process. read more
Researchers: Dr. Raphael Finkel, Dr, Gregory Stump (English)
We investigate the morphology of natural languages by using both generative and analytical tools. Generative tools include KATR, an extension of DATR for implementing default inheritance hierarchies, and a PFM (Paradigm Functional Morphology) Web site that allows the user to build and debug PFM theories. Analytical tools are centered around plats (charts of the paradigms of all inflection classes for a given language); these tools derive principal parts and related measures from the plats.
Decision-Theoretic Academic Advising
Researcher: Dr. Judy Goldsmith
During the course of a student's undergraduate education, many decisions are encountered which may impact short- and long-term academic success as well as relative enjoyment and (perceived) utility that are obtained by the student. Human advisors help the student advisee make decisions that can have positive major effects on their educational experience. The advisor's task is complicated by a potential lack of knowledge of the individual student's goals and preferences. Further, the potential long-term effects of actions may not be obvious even to experienced academic advisors.
In order to deal with the difficulties ecountered in academic advising our research group is developing tools and methods for generating stochastic models of an academic domain, and for fast stochastic planning and generation of advice. The project is divided into three areas: model construction, planning, and interface design. The academic domain poses challenges in each of these areas.
Probabilistic Computational Social Choice
One of the most common preference aggregation methods--the one most familiar to Americans--is election by majority. Other preference aggregation methods are not always recognized as such, for example, (sports) tournaments. One can view a sports tournament as an election where the best team wins. We can affect the outcome of a vote or tournament by voting and playing truthfully and to the best of our ability, etc, or by manipulating the aggregation process.
There are several methods by which aggregation schemes can be manipulated. The most intuitive and well known is by influencing individual agents (through payments or other means). In real-world systems, typically not everything (the influence, the vote, the result) is observable by the manipulator. With this project, we focus on uncertain outcomes: What happens if the manipulator has acces only to probabilities of agents' responses to attempts to influence them?
We achieve this through new model methods for established problems which take into account an agent's uncertainty about aspects of aggregation procedures. Once we have developed these new models, we study the complexity of lobbying and other influence methods in this uncertain world.
Web-based Interactive Organic Chemistry Homework
Researchers: Dr. Raphael Finkel, Dr. Robert Grossman (Chemistry)
ACE (Achieving Chemistry Excellence) is a web-based suite of programs for interactive chemistry homework. It follows the pedagogic principal that students learn best by working problems, but that they should not be shown a "correct" answer unless they have completely worked it out themselves. ACE provides a wealth of questions in Organic Chemistry that an instructor can assemble into assignments and examinations. The author of each question (perhaps the instructor of the course) provides a seriesof evaluators that check for aspects of the student response. The first evaluator that matches the response triggers both a score (on a 0-1 scale) and feedback. It the response is incorrect, the feedback aims to show the student in what way the response is wrong, not what the right answer might be. ACE currently has question types involving chemical structures, Lewis structures, mechanisms, multi-step synthesis, orbital energy diagrams, reaction coordinate diagrams, and more. Our current efforts are to expand the framework to include questions specific to other disciplines. read more
The video to the right provides a demonstration of ACE capabilities.
Researcher: Dr. Jane Hayes
The TraceLab Project seeks to develop an experimental workbench for designing, constructing, and executing traceability experiments, and facilitating the rigorour evaluation of different traceability techniques. TraceLab is similar in some respects to existing tools such as Weka, MatLab, or RapidMiner, except that it is highly customized to support rigorous Software Engineering experiments as opposed to general data mining ones. read more
The video below describes TraceLab and provides a short demonstration of its use.
Researcher: Dr. Jane Hayes
REquirements TRacing On target (RETRO.NET) is a tool developed by the software research group at UK to assist in the generation of traceability matrices between textual software engineering artifacts. For more information, contact Dr. Hayes at email@example.com.
The demonstration to the right provides a quick overview of the tool and its capabilities.
Efficient Algorithms for Functions with Infinitely-Many Variables
Reseacher: Dr. Grzegorz Wasilkowski
There are many computational problems dealing with functions of infinitely many variables. Such problems appear in, e.g., quantum chemistry and physics, financial mathematics, statistics, stochastic differential equations, and partial differential equations with random coefficients. Actually, many problems involving stochastic processes arre examples of such ∞-variable problems.
Currently available methods are very inefficient, and new algorithms need to be found. In this research project, we study a new family of methods, called Changing Dimension Algorithms. These algorithms allow us to approximate or integrate functions with infinitely many variables at the cost polynomial in ε, where ε is the error demand.