Improving Interpretability And Accuracy Of Artificial Intelligence In Natural Language And Image Understanding