See the blog post
Visualization is called Parallel Sets which are amazingly effective and efficient to show categorical variables in the dataset. I am using D3 parsets which is a parallel sets library on top of D3.
The visualization shows the distribution of the CS professors of their ranks, subfields, doctorate universities and university that they are working. In order to see the percentage of a subcategory (e.g. Full in the Rank category), hovering would suffice. You would be able to also see the percentage of the subcategory over all subcategories.
When you hover a path, you would be able to see the connections that create that particular path. (e.g: Full -> 2013 -> MIT) as well as the number of paths in the dataset and the percentage over all possible paths. You could sort by the size of each subcategory by hovering over the category(e.g. University) and then click on the size.