Interpreting Categorical Data Classifiers using Explanation-based Locality