Authorship Attribution Using Entropy

Abstract

Abstract We propose a new methodology for testing the authorship of a relatively small work compared with the large body of an author’s cannon. Our approach is based on comparing the entropy of the two samples. The difficulty lies in the fact that known estimators of entropy tend to have a large bias even when the sample size is fairly large. To deal with this, we suggest splitting the larger sample into several parts of length equal to the length of the smaller work. We then propose using these new sub-samples in a simple non-parametric test. We apply our methodology to test whether the poem “Shall I Die?” which is sometimes attributed to William Shakespeare was, in fact, written by him.