This project builds a standard, tertiary ROYGBV color wheel from thousands of images returned by queries for color terms. While the results mostly conform to expectations (e.g. 'orange' predominantly returns images of orange things), there are some important exceptions ('+blue +violet' yields alot of flesh-tone and is most definitely not safe for work).
Here, the interest lies in mapping a "semantic gradient" to a color gradient through the amoral & skewed lens of search engines.1
1. <Color> Wheel used bing.com as the search engine, although Google image search results overlap significantly with it. The choice of Bing is due to the Google api limit of 64 image downloads per search. Perhaps counterintuitively, Bing allows 1000 downloads per search.