New Algorithm Follows Human Intuition to Make Visual Captioning More Grounded