Although it would be easier for Stack Exchange Inc. to do this internally because they can query across stacks, any user could use Selenium to drive Chrome or Firefox to get a statistically sufficient sample of opened question titles on all the SE proposals and betas very slowly over a period of time so that impact on SE servers would be minuscule. The sample could be randomized by randomizing the date of the post from within the duration of the stack.
The data could be parsed into words and counted. Google's word frequency table could be used to normalize word significance so that words like 'ontological' would be more significant in evaluating topic overlap than 'the'. Then standard cross correlation between normalized word counts between the stacks could be used to create a Venn diagram. That might help people evaluate what overlap there really is between opened stacks.
There seems to be much talk about certain pairs of stacks and complete negligence of clear overlaps. It may also help pioneers develop their site name and description and drive their open-close policy with real information rather than conjecture and emotionally driven protectiveness of favorite stacks.
Has anyone done this work and is it posted anywhere as a mathematically clean and less disputable reference for meta discussion?