Real counts from the one-line descriptions of all 303 characters in the week 1 node file. Many of those lines are Wikipedia's boilerplate (…a fictional character appearing in American comic books published by Marvel Comics), so about a dozen words sit near 300 each.
Real counts from the full Wikipedia text of the same 303 characters (marvel_pages.zip on the course data page), tokenized the same way.
Each dot is one frequency. The bar to its right spans every word tied at that frequency, so the long flat run at 1 is all the words seen only once. The dashed line is an ideal Zipf curve with s = 1.