Authorship Attribution in Bengali Language

摘要

We describe Authorship Attribution of Bengali literary text. Our contributions include a new corpus of 3,000 passages written by three Bengali authors, an end-to-end system for authorship classification based on character n-grams, feature selection for authorship attribution, feature ranking and analysis, and learning curve to assess the relationship between amount of training data and test accuracy. We achieve state-of-the-art results on held-out dataset, thus indicating that lexical n-gram features are unarguably the best discriminators for authorship attribution of Bengali literary text.