USC Media Arts Honors Thesis Project

Once the information was organized in a digital file, I was able to explore several data visualization strategies. Before I knew I wanted to focus on geographic information, I first created a test visualization that shows word count per entry in my grandmother's diary from January 1986. Each column represents one entry and each dot in that row represents a word.

I went about making this visualization to figure out how I should structure my data within a JSON file so it can be parsed and visualized procedurally. Since the project involves so much hand transcribing of information, I needed to make sure that I wanted to use the D3.js library to do the heavy lifting for me in generating visualizations.

The visualization above looked at the frequency of times my grandmother mentioned a location in her journal, but it didn't give any sense of what that location looked like. For my second prototype, I wanted to viscerally connect the audience to the real locations my grandmother mentioned in her diary. For each address or street mentioned, I entered the location into Google Maps street view, zoomed out as far as I could and took a screenshot. After compiling all of these ../images, I used Adobe Photoshop's Batch Processing function to crop out all interface elements, resize the image to a manageable size and apply a black and white filter.

Next, I created a processing sketch that lets viewers interact with a timeline from January 1 - March 31, 1988. If my grandmother visited a location on the selected date, an image from Google Maps street view is shown. If she visited multiple locations, photos of each location are layered in the same window.

A third prototype during the exploration period of this project involved designing a system that compared the geographic metadata of individual entries to the textual information contained within each entry. To accomplish this, I created a new processing sketch with two modes. After a user selects an entry date, they can toggle between the non-redacted version of the date in question or the version redacted to just geographic information.

This prototype showed that many of the entries from my grandmother's diary do not reference specific locations. This discovery informed my decision to visualize location data in weeks and months, not days.

As I was working on the third prototype, I noticed that my grandmother’s entries often contained 10 digit sequences of numbers. After looking a few of them up, I uncovered that many of these were long distance telephone numbers to area codes outside of the state of Missouri. After getting a CSV with each date and phone number, I used Whitepages' Pro Graph API to map the phone numbers to specific addresses if possible and to zip codes if necessary. In the MapBox visualization, lines with lower opacity represent calls made only one time, while lines with higher opacity represent calls made multiple times.

The battery-powered control dial that users use to select the month being visualized connects to a Processing sketch over bluetooth low energy. The processing sketch talks to an LED display connected to an arduino board over serial and prints the month being visualized.

To create the grid on which the visualization will be displayed, I used an 8x8 grid of RGB LEDs. I attached foam board to the back of the shelf to easily mount and adjust the LEDs.

Each LED is connected to the matrix using a wire wrapping technique. I chose to wire wrap over solder each LED together because it elimnates the possibility of damaging or destroying LEDs. Additionally, using wire wrapping makes repairing easier because the wrapping can be undone quickly.

The rows and columns of the LED matrix interface with the Arduino Uno through a RGB matrix shield called a Colorduino. The Arduino is then connected to a MacBook running a processing sketch. This part of the project was heavily influenced by an open source project by @Lincomatic.

After wiring the LEDs together and connecting each strand to the Colorduino, I uploaded an open source program to check if each LED was working and addressable.