Color of words by Gene Kogan

This early work of machine learning art by the legendary Kogan explored the implicit colors associated with words and concepts. By running a set of words through a Google image search, Kogan analyzed the color distributions of the first hundred images that appeared. Using a self-organizing map (SOM) algorithm and a Gaussian mixture model (GMM), he synthesized these colors into a map of their visual representations. The project categorized the results into themes like seasons, biomes, holidays and politics. This quite early (pre-2014) instance of machine learning in creative expression highlights the lesser-known work of Kogan, how instrumental early image-classification attempts were and uncovers the hidden color palettes of language through data clustering and visualization techniques.