Typeface Map

Project Type: Data Visualization, Graphic Design

Computational Typeface Classification

After searching for ideas for things that hadn’t been tackled in the realm of computational information design yet, I discovered that the only visualizations of typeface classification and the relationships between different fonts had been static posters, so I thought this was a real opportunity to do something that hadn’t been done before. The concept has a fairly broad scope and can be expanded to include any number of ideas and applications, but for the sake of making it workable for a class project, I decided to focus my efforts mainly on the analysis of the letter shapes and the mapping in 2D space, where proximity represents a measure of similarity between two typefaces. The project had two deliverables: a large poster representing the map, as well as an interactive applet implemented in Processing and Java.

Many people asked me, “Why do this project? Why are you interested in this?” The answer is that I think a lot of us graphic designers carry a vague notion of this typeface map in our heads, but if you asked us exactly to describe it (or even draw it) I think we’d have a hard time just because I think our understanding of the relationships between typefaces is based as much as, if not moreso, on this intuitive sense, rather than facts and math. So, I was interested in comparing the maps in our heads with a mapping based on mathematical analysis.

The computational aspect of the project had two main components: Analysis of the letter shapes using various metrics and shape analysis methods, and mapping the nodes in 2D space using PCA. The letter shape analysis was done using the Geomerative library for Processing, while the mapping using PCA used the TuftsGeometry library for Java. Once I figured out how to do the PCA projection using this library, I just needed to add any new variables to the matrix input each time, so that was fairly simple. However, since the variables didn't always have the same units of measure (eg. some were ratios), finding the correct weightings took up a lot of time in this project.

For the sake of comparing the mapping to our learned knowledge related to typefaces, the program also reads in a CSV file with some information for each typeface such as the year it was made and the pedantic classification it’s been given by ATypI. The digital applet allows the user to overlay this information on the mapping to see if any interesting and/or unexpected results are shown.