Publication Date

2013

Document Type

Thesis

Committee Members

Lang Hong (Committee Member), Jack Jean (Committee Member), Thomas Wischgoll (Advisor)

Degree Name

Master of Science in Computer Engineering (MSCE)

Abstract

General Recognition Theory (GRT) is a multidimensional signal detection theory framework for capturing sources of perceptual and decisional dependence. The primary type of data for GRT models is an identification-confusion matrix derived in a complete factorial identification task. This confusion matrix plots the responses of study participant for a given signal. The responses may reveal that participants were unable to recognize the signal properly. Such violations of any type of independence in the GRT framework result in response patterns that reflect some form of correlation in the GRT space. While an individual confusion matrix is rather small and relatively easy to visualize, similar studies may be repeated a lot of times resulting in thousands if not millions of confusion matrices that have to be visualized. This thesis describes methods that adapt the web-based D3 visualization framework combined with pre-processing tools for the raw data to identify ways of visualization methodologies that enable domain specialists to more easily interpret their data. As the D3 framework utilizes Javascript and scalable vector graphics (SVG) to generate the visualizations it can run readily within the web browser to directly enable deployment of the visualization algorithms by the domain specialists. Parallel coordinate plots and heat maps were developed for the confusion matrix data, and the results were shown to a GRT expert for an informal evaluation of their utility. A second part is the visualization of database for semantic cluster which come from DBpedia. DBpedia is a crowd-sourced community effort to extract structured information from Wikipedia. Its database is served as Linked Data on the Wikipedia. VTK as a generic C++ class library has a wide range of scientific and engineering usage. It can be used with another C++ based virtual environment library VRUI (virtual reality user interface). VTK and VRUI based semantic cluster database visualizations with different layout approaches were developed. There is a clear benefit to model interpretation from these visualizations when researchers need to interpret larger amounts of simulated data.

Page Count

77

Department or Program

Department of Computer Science and Engineering

Year Degree Awarded

2013


Share

COinS