1. 4
  1.  

  2. 2

    There was some super interesting related work to this at SIGGRAPH a couple years ago as well:

    http://web.engr.oregonstate.edu/~mjb/cs550/Projects/Papers/CSemanticShapeEditing.pdf

    It seems like the general state of semantic editing is that it works in some very specific conditions but when those conditions are met it works unbelievably well.