Browsing by Author "Robert St. Amant, Committee Member"

Now showing 1 - 19 of 19

An Accessible Cognitive Modeling Tool for Evaluation of Human-Automation Interaction in the Systems Design Process.
(2010-11-04) Gil, Guk Ho; David Kaber, Committee Chair; Robert St. Amant, Committee Member; Michael Feary, Committee Member; David Dickey, Committee Member; Yuan-Shin Lee, Committee Member
Automated Scaffolding of Task-Based Learning in Non-Linear Game Environments.
(2011-01-05) Thomas, James; Robert Young, Committee Chair; Patrick Fitzgerald, Committee Member; Robert St. Amant, Committee Member; Jon Doyle, Committee Member
Bowyer: A Planning Tool for Bridging the Gap between Declarative and Procedural Domains
(2008-02-03) Cash, Steven Patrick; R. Michael Young, Committee Chair; Robert St. Amant, Committee Member; James C. Lester, Committee Member
Cognitive Models of Discourse Comprehension for Narrative Generation
(2009-07-27) Niehaus, James Michael; James Lester, Committee Member; Stephen Mitroff, Committee Member; Robert St. Amant, Committee Member; Jon Doyle, Committee Member; R. Michael Young, Committee Chair
Recent work in the area of narrative generation has sought to develop systems that automatically produce experiences for a user that are understood as stories. Much of this prior work, however, has focused on the structural aspects of narrative rather than the process of narrative comprehension undertaken by readers. Cognitive theories of narrative discourse comprehension define explicit models of a reader's mental state during reading. These cognitive models are created to test hypotheses and explain empirical results about the comprehension processes of readers. They do not often contain sufficient precision for implementation on a computer, and thus, they are not yet suitable for computational generation purposes. This dissertation employs cognitive models of narrative discourse comprehension to define an explicit computational model of a reader's comprehension process during reading, predicting aspects of narrative focus and inferencing with precision. This computational model is employed in a narrative discourse generation system to select content from an event log, creating discourses that satisfy comprehension criteria. The results of three experiments are presented and discussed, exhibiting empirical support for the computational reader model and the results of generation. This dissertation makes a number of contributions that advance the state-of-the-art in narrative discourse generation: a formal model of narrative focus, a formal model of online inferencing in narrative, a method of selecting narrative discourse content to satisfy comprehension criteria, and implementation and evaluation of these models.
Cognitive Task Analyses for Life Science Automation Training Program Design.
(2008-08-11) Green, Rebecca; Regina Stoll, Committee Member; Robert St. Amant, Committee Member; Christopher Mayhorn, Committee Co-Chair; David Kaber, Committee Chair
The purpose of this study was to develop a systematic approach to the translation of Cognitive Task Analyses (CTAs), including Goal Directed Task Analysis (GDTA) and Abstraction Hierarchy (AH) models, into a Situation Awareness (SA) based training program for operators of High-throughput (biological) screening (HTS) systems. Traditional on-the-job (OTJ) training of new HTS operators usually consists of several weeks of assisting a lead biochemist to become familiar with methods and automated systems. Unfortunately, this approach to training is typically unstructured and learning results may be highly variable. In order to design instruction to support learning of cognitive processes as part of HTS, the information demands engendered by the task need to be identified. This can be achieved using CTAs as the basis for training program design. Various CTA methods, including the Critical Decision Method (CDM) and Precursor-Action-Results-Interpretation, have been used to develop training. However, no standardized methods exist for relating the outcomes of the integration of multiple CTA methods to support training program design. This study, therefore, combined information requirements from a GDTA and system resource requirements identified through AH models to establish content on HTS processes for delivery through an electronic training program. The goals and sequences of task steps within the training program were identified by the GDTA. The use of AH models of the HTS system provided a method for determining the purpose and function of the software and devices relative to different operator functional requirements. This combination of information from the CTAs provided a systematic approach for specifying training strategies and parameters. The training program presented learners with content for development of the three levels of operator SA (perception, comprehension, and projection) and knowledge structures pertaining to HTS system operations. Following development of the prototype electronic training program and the comparison traditional training program, an evaluation occurred through a three-part survey with comparison to the traditional lab training provided to expert operators of an HTS system. The evaluation incorporated two knowledge assessment tests, a usability survey, and a survey of the effectiveness of the SA elements of the training program. Results provided preliminary evidence that a CTA-based training program can improve operators' knowledge structures beyond OTJ training. Furthermore, operator performance on SA questions indicated improvements in knowledge structures associated with perceptual elements, comprehension of those elements, and projection of the future states of HTS systems. Additionally, since experience can lead to differences in operator mental models pertaining to HTS systems, the effect of two types of overall experience and individual task experience were measured. Results indicated that the CTA-based training program was effective in providing improved SA knowledge and general knowledge structures for HTS operators beyond their initial knowledge of the system (i.e., considering work experience and education). A heuristic-based evaluation of both training programs identified few unique usability problems, suggesting the usability of the training programs did not interfere with the development of learner knowledge structures. Finally, on the basis of these results, a set of general guidelines for the design of the CTA-based training programs was developed. These guidelines included methods for structuring the components of the training program to support the three levels of SA and the amount of text that should be shown for each task.
Commitment-Based Business Process Modelling and Enactment
(2005-04-03) Wan, Feng; Robert St. Amant, Committee Member; Peter Wurman, Committee Member; Munindar P. Singh, Committee Chair; James Lester, Committee Member
Business process management faces challenges in dealing with business abnormalities and ever-changing business requirements. Traditional business process management approaches evolved from software engineering and workflow management where activities, messages and control logic are given prominence. The resulting models specify low-level details of execution and coordination. However, difficulties arise when modelling long-lived business transactions involving information updates and execution exceptions. To handle such situations, current approaches implement excessive activities without suitable abstractions, thereby arbitrarily fragmenting the business requirements. We propose a commitment-based approach for business process modelling that formulates business processes as multiagent systems. Organizational structure and its effect on interactions are described using commitments and causality. Agents act as process executors and maintain the commitments made to each other. Updates and exceptions yield commitment operations under which processes are updated and reexecuted. Our approach brings commitment semantics into business modelling and enables agent collaboration for business process enactment. We derive commitment protocols from agent conversations and generate agent execution models. We also formalize our approach using the Π-calculus and prove its correctness. To demonstrate the practical use of our approach, we formalize multiparty agreements with commitments and present algorithms on how to detect agreement conflicts and build satisfiable commitment sets.
Decision-Theoretic Narrative Planning for Guided Exploratory Learning Environments
(2006-07-27) Mott, Bradford Wayne; Patrick J. FitzGerald, Committee Member; James C. Lester, Committee Co-Chair; R. Michael Young, Committee Co-Chair; Robert St. Amant, Committee Member
Interactive narrative environments have been the focus of increasing attention in recent years. A key challenge posed by these environments is narrative planning, in which a director agent orchestrates all of the events in an interactive virtual world. To create effective interactions, the director agent must cope with the task's inherent uncertainty, including uncertainty about the user's intentions. Moreover, director agents must be efficient so they can operate in real time. To address these issues, we present U-DIRECTOR, a decision-theoretic narrative planning architecture that dynamically models narrative objectives (e.g., plot progress, narrative flow), storyworld state (e.g., physical state, plot focus), and user state (e.g., goals, beliefs) with a dynamic decision network (DDN) that continually selects storyworld actions to maximize narrative utility on an ongoing basis. DDNs extend decision networks by introducing the ability to model attributes whose values change over time; decision networks extend Bayesian networks by supporting utility-based rational decision making. The U-DIRECTOR architecture also employs an n-gram goal recognition model that exploits knowledge of narrative structure to recognize users' goals and an HTN planner that operates in two coordinated planning spaces to integrate narrative and tutorial planning. U-DIRECTOR has been implemented in a narrative planner for an interactive narrative learning environment in the domain of microbiology in which a user plays the role of a medical detective solving a science mystery. Formal evaluations suggest that the U-DIRECTOR architecture satisfies the real-time constraints of interactive narrative environments and creates engaging experiences.
Design and Prototyping of a Cognitive Model-based Decision Support Tool for Anesthesia Provider Management of Crisis Situations
(2006-08-21) Segall, Noa; Regina Stoll, Committee Member; Gary Mirka, Committee Member; Christopher Mayhorn, Committee Member; Melanie Wright, Committee Member; Robert St. Amant, Committee Member; David B. Kaber, Committee Chair
This research involved the prototyping of a decision support tool (expert system) for use by anesthetists in crisis situations, in order to promote prompt and accurate patient diagnosis, care, and safety. The tool alerts anesthetists to a developing crisis, manifested by changes in certain patient physiological variables, and provides them with a list of potential causes and preventive measures for dealing with the crisis. The tool provides advice in an unobtrusive manner. Information is presented in a format requiring minimal interaction with the system interface. Decision support tools for managing patient crisis situations may be useful in large hospitals where an attending anesthesiologist supervises multiple nurse anesthetists or anesthesiology residents that are delivering drugs to patients across operating rooms. Such a tool can provide support to nurses and residents when the attending physician is not present, and can warn of potential crisis situations that would prompt the anesthesia provider to contact an attending physician. The attending physician may also use the tool as a quick method of learning patient status when entering an OR. In addition, the tool could be used by practitioners working alone to deliver anesthesia. A novel approach was applied to the development of the decision support tool to support anesthesiology decision-making. First, a hierarchical task analysis was conducted to identify the procedures of the anesthetist in detecting, diagnosing, and treating a critical incident, specifically, myocardial infarction. Second, a cognitive task analysis was carried out to elicit the necessary goals, decisions, and information requirements of anesthetists during crisis management procedures. The results of these analyses were then used as bases for coding a cognitive model using GOMS (goals, operators, methods, selection rules), a high-level cognitive modeling language. EGLEAN (error-extended GOMS language evaluation and analysis tool), an integrated modeling environment, was used as a platform for developing and compiling the GOMS model and applying it to a Java-based simulation of a patient status display. After the anesthetist's decision-making process was captured in GOMS, a basic interface for the decision support tool was prototyped (extending traditional OR displays) to present output from the computational cognitive model by using ecological interface design principles. Finally, a preliminary validation of the tool and interface (patient state and cognitive model output displays) was performed with samples of expert anesthesiologists and human factors professionals in order to assess the usability and applicability of the decision support tool. The anesthesiologists indicated that they would use the decision support tool in crisis situations and would recommend its use by junior anesthesia providers. The human factors experts provided comments on the interface's compliance with usability principles, such as providing prompt feedback and preventing errors. This research has provided insight into anesthetist decision-making processes in crisis management. It resulted in a prototype of a cognitive model-based decision support tool to augment anesthetist decision-making abilities in these situations.
Development of a Haptic-based Rey-Osterrieth Complex Figure Testing and Training System with Computer Scoring and Force-feedback Rehabilitation Functions.
(2010-05-06) Li, Yingjie; David Kaber, Committee Chair; Yuan-Shin Lee, Committee Chair; Christopher Healey, Committee Member; Simon Hsiang, Committee Member; Larry Tupler, Committee Member; Robert St. Amant, Committee Member
Examining and Explaining the Effects of Non-Iconic Conformal Features in Advanced Head-up Displays on Pilot Performance
(2009-04-15) Kim, Sang-Hwan; David B. Kaber, Committee Chair; Robert St. Amant, Committee Member; Eric Wiebe, Committee Member; Brad Mehlenbacher, Committee Member; Nancy Currie, Committee Member
The primary objective of this study was to assess the impact of Synthetic Vision System (SVS) and Enhanced Vision System (EVS) depictions of terrain features on pilot performance when displayed in an advanced head-up display (HUD) during various phases of a landing approach under instrument meteorological conditions (IMCs). SVS is a display system that presents terrain features using a wireframe grid rendered polygons by integrating terrain databases with a global positioning system. EVS displays present an actual out-of-cockpit view using a forward looking infrared camera. In the experiment as part of this study, video stimuli presenting varied HUD configurations were pre-recorded using a high-fidelity flight simulator at NASA Langley and presented to eight pilots later in a lab environment. The HUD videos from the high-fidelity simulator were combined with out-of-cockpit views from a lab simulator. The flight scenario consisted of an approach and landing on a runway (Reno, Nevada International Airport (KRNO), 16R (right)) under IMC. Each pilot completed eight trials based on a within-subjects experimental design and one additional trial to collect verbal protocols on specific display feature use. The independent variables included four display configurations (baseline, SVS-only, EVS-only, and a combination of SVS and EVS features) and two visibility conditions (IMC-day versus IMC-night). Every display configuration included tunnel features (highway-in-the sky) showing the designated flight path. The experiment involved observing pilot performance in four segments during the approach and landing. Dependent variables included flight path control performance, pilot SA, workload, and subjective preferences. Flight path control performance was determined based on pilot errors in tracking a flight path marker in the pre-recorded videos with a super-imposed cursor using test pilots yoke controls. Pilot situation awareness (SA) was measured using SAGAT (the Situation Awareness Global Assessment Technique) in order to evaluate pilot perception, comprehension, and projection for three types of pilot SA (spatial, system, and task awareness). Workload measures were recorded using the NASA-TLX (Task Load Index) and heart-rate. In order to develop explanations of pilot behavior under the various HUD conditions, a video record of the additional test trial was reviewed by each subject using a verbal protocol analysis and semi-structured interview. Results revealed SVS to support overall pilot SA but to degrade flight path control performance due to confusion of visual features, EVS caused pilots to focus on path control but decreased System awareness because of visual distractions of some imagery. The combination of SVS and EVS features generated offsetting effects; however there were decrements in performance in the final landing phase due to clutter effects. In general, display configurations did not affect spatial awareness but pilot awareness of system information was impacted. The IMC-day condition produced worse flight performance than night flight due to the low visual saliency of HUD imagery in daylight. Flight performance was not different among phases of flight but different levels and types of pilot SA were affected by segment. Because the main task in the study was the tracking task, results did not reveal differences of conditions in terms of workload measures. Interestingly, patterns of pilot preference for displays did not match with the results of objective performance and SA measures. Pilots gave higher ratings of SA support and safety for the SVS and EVS displays with the lowest ratings going to the combination. Ratings on annoyance increased with increases in display visual content. The verbal protocol analysis yielded sequential and non-sequential lists of pilot tasks and behaviors and critical pilot comments. The analysis also identified the required information and alternative methods of performance for specific flight tasks in the scenario. This analysis was used to explain the experimental results and describe pilot behaviors with the SVS and EVS displays in the flight scenario. This study assessed advanced HUD feature effects on pilot performance, using an elaborate SAGAT method for measuring pilot SA, and developed a CTA for interpreting experimental results. Further studies need to be conducted to evaluate the advanced HUDs under various flight situations using a more realistic flight simulator as a basis for optimal design. In addition, cognitive model of pilot behavior based on CTA needs to be developed for predicting performance and SA implications of HUD design.
Explorations in Three-Dimensional User Interfaces for Learning Environments
(2003-11-07) Casstevens, Randy Mark; Robert St. Amant, Committee Member; James Lester, Committee Chair; Patrick FitzGerald, Committee Member
Computerized learning environments have the potential to dramatically improve the pedagogical effectiveness of the educational process. The computer will not replace the teacher in the classroom, but it could play a significant role in the students' education. The computerized learning environment could provide each student an interactive, custom lesson. This thesis examines how to develop a three-dimensional user interface that would improve the learning environment. We draw examples from two three-dimensional learning environments; the Steve and PhysViz projects. This thesis discusses four stages of the software development cycle (analysis, design, prototyping, and evaluation) to consider when developing a three-dimensional user interface. We first describe the potential characteristics of a learning environment that benefit from a three-dimensional interface. Next we explore the use of interaction metaphors and affordances in a three-dimensional learning environment. We found that direct manipulation of the interface can be very useful for a learning environment and also saw how this can be facilitated by affordances. After the design considerations, we begin examining issues that arise when prototyping a three-dimensional learning environment. Our discussion focuses on issues we encountered with Java 3D when implementing our three-dimensional world for the PhysViz project. We also introduce some ideas about camera control, navigation of the student, and the display of text. Finally, we propose an evaluation plan for three-dimensional user interfaces for learning environments. This thesis provides software developers of learning environments with a guide to the advantages and disadvantages of using a three-dimensional interface. From developing PhysViz, a physics tutorial application, we found that a three-dimensional interface was beneficial. The additional dimension added to the richness of the interface and improved the pedagogical effectiveness of our learning environment.
Hierarchical Charged Particle Filter for Multiple Target Tracking.
(2010-11-10) Bhatia, Amit; Griff Bilbro, Committee Chair; Wesley Snyder, Committee Chair; Christopher Healey, Committee Member; Robert St. Amant, Committee Member
Interest-Matching Comparisons using CP-nets
(2007-04-03) Wicker, Andrew White; Jon Doyle, Committee Chair; Robert St. Amant, Committee Member; Peter Wurman, Committee Member
The formation of internet-based social networks has revived research on traditional social network models as well as interest-matching, or match-making, systems. In order to automate or augment the process of interest-matching, we follow the trend of qualitative decision theory by using qualitative preference information to represent a user's interests. In particular, a common form of preference statements for humans is used as the motivating factor in the formalization of ceteris paribus preference semantics. This type of preference information led to the development of conditional preference networks (CP-nets). This thesis presents a method for the comparison of CP-net preference orderings which allows one to determine a shared interest level between agents. Empirical results suggest that distance measure for preference orderings represented as CP-nets is an effective method for determining shared interest levels. Furthermore, it is shown that differences in the CP-net structure correspond to differences in the shared interest levels which are consistent with intuition. A generalized Kemeny and Snell axiomatic approach for distance measure of strict partial orderings is used as the foundation on which the interest-matching comparisons are based.
A Knowledge Maturity Model: An Integration of Problem Framing, Software Design, and Cognitive Engineering
(2004-03-19) Markow, Tanya Thais; Thomas L. Honeycutt, Committee Chair; Christopher G. Healey, Committee Member; Robert St. Amant, Committee Member
The Knowledge Maturity Model (KMM) is a new model proposed as an alternative to an existing software engineering evaluation model, the Capability Maturity Model (CMM). The KMM is offered as a solution to some key weaknesses of the CMM. The CMM was developed in the early 1980s, when highly structured programming and business practices were the standard. In the current agile methods computer science environment, it is often difficult to evaluate a company which employs agile methods using the CMM methodology. The CMM consists of five levels; in order to claim the next higher level, all tasks of that level must be accomplished. Many companies operating with agile software engineering and management practices tend to be performing at many levels within the CMM, making it difficult to assign such an organization an appropriate CMM level designation. The KMM proposes instead an evaluation of the actual inner processes the company uses to develop software, rather than its ability to achieve a given set of tasks, as required by CMM. It will be shown that the KMM bridges the gap between the CMM and agile methods by employing the Knowledge Insight Model (KIM). The KIM is an iterative process that employs four key roles: Framer, Maker, Finder and Sharer. The Framer is responsible for the 'big picture' of project management, including defining requirements and scope. The Maker must create new concepts and code for solving the problem. The Finder seeks out existing knowledge and information to help solve the problem. The Sharer must create and maintain a database of the project and ensure that all involved get the information they need. The Knowledge Maturity Model incorporates the concept of levels or states of maturity from the CMM, and the core fundamentals of the KIM: the roles and an iterative process. The synergy of these concepts gives rise to the four state model of the KMM: recognition and use of the Plan, Do, Check, Act cycle, use of the four roles of KIM, use of an iterative process, and finally, the fully working inner mechanism or sharing mechanism of the KIM. The KMM allows an organization to choose any traditional software engineering methodology for a given project by providing the roles-based structure to make shifting between software engineering methodologies easier, allowing companies to tailor their process for specific projects. KMM ties together the three fundamental centers that comprise the process of developing software: systems engineering, software engineering, and cognitive engineering. The KMM solves the systems engineering problem by providing a generalized process that is a superset of any given software engineering methodology. Because KMM provides a superset to all existing software engineering methodologies, it frees up an organization to choose the one that best suits a given project, rather than always having to use one standard approach, therefore addressing the software engineering aspect. At the heart of KMM are the four roles, which addresses the need to completely incorporate people into the process, thus bringing in the cognitive engineering side of the discipline.
Pairwise Document Similarity using an Incremental Approach to TF-IDF.
(2010-08-09) Venkatesh, Jayashree; Christopher Healey, Committee Chair; Robert St. Amant, Committee Member; Jon Doyle, Committee Member
Perception Driven Search Strategies For Effective Multi-Dimensional Visualization
(2003-02-13) Kocherlakota, Sarat Mohan; Thomas L. Honeycutt, Committee Member; Robert St. Amant, Committee Member; Christopher G. Healey, Committee Chair
Tracking and analysing large amounts of information in many different application areas is a critical problem. One approach to address this problem, is the use of multi-dimensional visualizations to represent large datasets. Visualizations can be constructed effectively by the use of visual features and properties like color and texture. Our objective is to construct multi-dimensional visualizations using perceptually salient visual features which support rapid visual analysis and exploration of large datasets. We use a visualization system called ViA use to construct effective visualizations. We present a search technique incorporated in ViA, that finds effective attribute-feature mappings to represent multi-dimensional datasets in a perceptually salient fashion. ViA evaluates the salience of attribute-feature mappings using evaluation engines. These evaluation engines also suggest hints that recommend how the mapping can be improved perceptually. The search technique we developed, uses dataset properties, and the hints generated by the evaluation engines to quickly and efficiently produce perceptually salient mappings. Perceptual guidlines were established from studies and experiments on human perception. ViA works as a semi-automated visualization system that uses effective search technique to find salient mappings. Applying ViA to practical datasets indeed proves the effectiveness of ViA. We think ViA can also produce salient visualizations in a variety domain areas since the guidelines for generation of effective visualizations are based on human perception.
A Systematic Model Building Process for Predicting Actionable Static Analysis Alerts
(2009-07-07) Heckman, Sarah Smith; Steffen Heber, Committee Member; Laurie Williams, Committee Chair; Tao Xie, Committee Member; Robert St. Amant, Committee Member
Automated static analysis tools can identify potential source code anomalies, like null pointers, buffer overflows, and unclosed streams that could lead to field failures. These anomalies, which we call alerts, require inspection by a developer to determine if the alert is important enough to fix. Actionable alert identification techniques can supplement automated static analysis tools by classifying or prioritizing the alerts generated by automated static analysis such that the likelihood of a developer inspecting actionable alerts first is increased. By classifying and prioritizing actionable static analysis alerts, the developer will focus his or her time on inspecting and fixing actionable alerts rather than inspecting and suppressing unactionable alerts. The goal of my research is to reduce inspection time by accurately predicting actionable and unactionable alerts when using static analysis by creating and validating a systematic actionable alert identification model. The Systematic Actionable Alert Identification (SAAI) process uses machine learning to identify actionable alerts. Investigation of the following three hypotheses will inform the goal of my research: - Hypothesis 1: The artifact characteristics of an alert and the surrounding source code are predictive of the actionability of an alert. - Hypothesis 2: A systematic actionable alert identification technique using machine learning can accurately identify actionable alerts. - Hypothesis 3: A systematic actionable alert identification technique using machine learning is project specific. A benchmark, FAULTBENCH, provides the evaluation framework for the proposed SAAI model building process and comparison with other actionable alert identification techniques. The dissertation presents a feasibility study and three empirical studies evaluating the hypotheses above. The feasibility study evaluates an adaptive actionable alert identification technique that utilizes the alertÃ¢â‚¬â„¢s type and code location in addition to developer feedback to prioritize actionable alerts. The first empirical study investigates hypotheses 1-3 using FAULTBENCH on 15 SAAI models generated on five treatments for each of three subject programs. The treatments considered different grouping of alerts within revisions to train and test SAAI. The second empirical study is a comparative evaluation of the generated SAAI models with other actionable alert identification techniques in further evaluation of Hypothesis 2. Additionally, an empirical user study was conducted where students in the senior capstone project course used a custom SAAI model during development of their software project. Selection of predictive artifact characteristics as part of the SAAI process suggests the acceptance of hypothesis 1. All but four of the 58 artifact characteristics used to build SAAI models were in one or more of the artifact characteristics subsets. The SAAI model identified actionable and unactionable alerts with greater than 90% accuracy for eight of the 15 FAULTBENCH subject treatments. Comparing SAAI models with other actionable alert identification techniques from literature found that SAAI models had the highest accuracy for 11 of the 15 treatments when classifying the full alert sets. Both of the above results support hypothesis 2. Due to accuracies greater than 90% when applying artifact characteristic subsets and machine learning algorithms for one subject program to another subject program, hypothesis 3 is not supported on the evaluated subject programs. The contributions of this work are as follows: - A systematic actionable alert identification model building process to predict actionable and unactionable automated static analysis alerts; - A benchmark, FAULTBENCH, for evaluating and comparing actionable alert identification techniques; and - A comparative evaluation of systematic actionable alert identification models with other actionable alert identification techniques from literature.
Tools for Business Protocols.
(2010-05-05) Narayan Rajagopalan, Koushik; Munindar Singh, Committee Chair; Robert St. Amant, Committee Member; Kemafor Ogan, Committee Member
Visualization of Large Document Collections.
(2010-11-03) Hsiao, Ping-Lin; Christopher Healey, Committee Chair; Robert St. Amant, Committee Member; Benjamin Watson, Committee Member; James Lester II, Committee Member