Mapping Twitter before the presidential elections in France, 2022
I visualized the French political Twitter just before the presidential election of 2022, in collaboration with other researchers and the journalists of the French newspaper Le Monde.
The project includes a map for a double-page print, images for offline and online articles, and a high resolution rendering published online as a zoomable map.
The network map represents about 30,000 Twitter accounts derived from an original corpus of about 600,000 tweets mentioning at least one candidate to the presidential election between the 3rd and the 21st of March 2022.
The map is the product of a collective work initiated by Guilhem Fouétillou, Chief Evangelist Officer and cofounder at Linkfluence, a Meltwater Company. Guilhem teamed up with researchers and journalists from Le Monde to analyze the data he had harvested on Twitter. Our collective gathered them from the end of January and until the publication of a series of articles in the digital and paper editions of March 31st, April 1st and 2nd, 2022.
Network layout
Although the links are not display, the visualization is a network. The links are implicitly represented by node positions, as the layout algorithm used to place them tries to bring densely connected groups together. The network was cleaned to optimize its readability.
Colors
Only certain nodes have a color, because the journalists manually curated the political affiliation of only the most prominent Twitter account (it is time consuming). The color of those nodes is allowed to slightly diffuse around them, which makes the clusters of the main political parties more visible.
Hillshading
The local node density is highlighted with hillshading and a hypsometric gradient. It was inspired by 1960's maps of the Moon by the NASA.
Labels
The map is isotropic: it has no favorite direction. Likewise, the labels flow like rivers from high-density areas to low-density areas, but without curving too much. They do not overlap and keep a readable size. The presidential candidates have a bolder font to allow the reader find them more easily.
Labels need a border, or readability issues arise:
But the background color depends on the hypsometric gradient and the spilling node colors. A uniform color gets too busy:
The labels border need to be composited in a sophisticated way to feel natural.
Variants
The main map was declined at different formats for different needs.