LSTM language model and classifier trained on a Sub-Word Level
Bengali Authorship Attribution through ULMFit trained on a Sub-Word Level.
Bengali Words have common roots.
Breaking the words into sub-word level increases the language model's ability to handle rare words better.