How might we visualize the propensity of a letter to be part of a typo?