Authorship categorization of public domain literature

We defined a set of quantifiable features for authorship categorization. We performed our experiments on public domain literature - all books analyzed were obtained in plain text format through Project Gutenberg's online repository of classic books. We tested three machine learning algorithms: Artificial Neural Network, Naïve Bayes Classifier, and Support Vector Machine with our features. We found that certain features, such as punctuation and various suffixes result in a higher accuracy. In addition, the Support Vector Machine classifier produces repeatedly higher accuracies than other classifiers and seems to be a far superior method of classification in terms of authorship categorization. © 2016 IEEE.

1 citations (Crossref) [2023-10-31] tex.type: Proceedings paper