In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
will defend his dissertation
SubjectBook: Data Management and Visualization Methods for Affective Studies
Managing affective data sets is very challenging for two main reasons. First, their life cycles consist of a series of cumbersome and time-consuming activities performed by different people. Second, investigators are increasingly overwhelmed by the size and complexity of data generated by affective studies. Such studies are longitudinal, and feature multimodal data, such as psychometric scores, imaging sequences, and signals from wearable sensors, with the latter streaming continuously for hours on end. The lack of tools for helping researchers manage their studies diminishes their ability to finish collecting, analyzing, and sharing affective data sets within reasonable amounts of time and effort. Moreover, it is difficult to avoid human errors if some tasks, such as data collection and curation, are performed manually. Importantly, some critical tasks, such as quality assurance and exploratory data analysis, can not be performed efficiently unless using appropriate representations for presenting and displaying relationships among collected data. In this work, we introduce SubjectBook, an integrated tool for managing affective studies throughout their life cycles, from designing the experiments to analyzing and sharing the generated data. In this tool, data collection and curation phases have been automated and validated. This enables researchers to have access to their own data in real-time. Additionally, meaningful visual representations of data are provided. Various tools that were proposed to tackle this problem provide visualizations of the original data only; they do not support higher level abstractions. Uniquely, SubjectBook operates at three levels of abstraction, mirroring the stages of quantitative analysis in hypothesis-driven research. The top level uses a grid visualization to show the study's significant outcomes across subjects. The middle level summarizes, for each subject, context information along with the explanatory and response measurements in a construct reminiscent of an ID card. This enables the analyst to appreciate within subject phenomena. Finally, the bottom level brings together detailed information concerning the inner and outer state of human subjects along with their real-world interactions - a visualization fusion that supports cause and effect reasoning at the experimental session level. SubjectBook was evaluated on a case study focused on driving behaviors.
Date: Friday, November 11, 2016
Time: 1:00 PM
Place: HBSC 350
Advisor: Dr. Ioannis Pavlidis
Faculty, students, and the general public are invited.