Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Electrical and Computer Engineering

Committee Chair

David S. Ebert

Committee Member 1

Ross Maciejewski

Committee Member 2

Arif Ghafoor

Committee Member 3

Edward J. Delp


Textual labels are an essential part of most visualizations used in practice. However, these textual labels are mainly used to annotate other visualizations rather than being a central part of the visualization. Visualization researchers in areas like cartography and geovisualization have studied the combination of graphical features and textual labels to generate map based visualizations, but textual labels alone are not the primary focus in these representations. The idea of using symbols in visual representations and their interpretation as a quantity is gaining more traction. These types of representations are not only aesthetically appealing but also present new possibilities of encoding data. Such scenarios regularly arise while designing visual representations, where designers have to investigate feasibility of encoding information using symbols alone especially textual labels but the lack of readily available automated tools, and design guidelines makes it prohibitively expensive to experiment with such visualization designs. In order to address such challenges, this thesis presents the design and development of visual representations consisting entirely of text. These visual representations open up the possibility of encoding different types of spatial and temporal datasets. We report our results through two novel visualizations: typographic maps and text-based TextRiver visualization. Typographic maps merge text and spatial data into a visual representation where text alone forms the graphical features, mimicking the practices of human map makers. We also introduce methods to combine our automatic typographic maps technique with spatial datasets to generate thema-typographic maps where the properties of individual characters in the map are modified based on the underlying spatial data. Our TextRiver visualization is composed of collection of stream-like shapes consisting entirely of text where each stream represents thematic strength variations over time within a corpus. Such visualization enables additional ways to encode information contained in temporal datasets by modifying text attributes. We also conducted a usability evaluation to assess the potential value of our text-based TextRiver design.