The semantics of the U.S. street map

This poster reappropriates tools of textual analysis into spatial humanities research by treating the road network of the United States as a single, massive text (of street names) read by navigators every day, and applying the increasingly common steps of representation learning to them; a high dimensional embedding using word2vec that captures forms of inter-relation between names, and a low dimensional re-embedding using UMAP for visualization.The resulting visualization opens a useful space between the spatial and textual humanities that looks to see how the evolving patterns of distant reading may be useful in thinking about the landscape as a text in the most literal way possible.