This could be a good machine learning project. What challenges do you sense at the first glance?
Discreet industries have quite specific legal terminology (in my experience) and machine learning is probably better suited to weeding out mistakes in documents rather than interpreting the contents within them. It's the interpretation that earns the big money. The example you've given is useful - but useful to an audience with specific prior knowledge and an existing familiarity with the licence landscape. Expanding beyond this is probably going to be an exercise in unintended consequences.
The idea has a lot of merit and would be truly disruptive it it could be achieved, but if lawyer jokes have taught us anything, there is no shortage of motivation to reduce the worlds reliance on lawyers, but nobody has yet come up with a good solution (to my knowledge).