Authorship Attribution on Bengali Literature using Stylometric Features and Neural Network

Abstract

Every writer has his/her personal writing style. In the era of technology, authorship attribution is a big problem in natural language processing because fake writers can publish other writers' contents and it is difficult to identify the real author. Various kinds of features such as frequently used words, word length, sentence length, WH words, Number, etc. were analyzed to identify and specify a writer’s writing style. A statistical analysis of different articles by different writers was created that can identify the real author. An artificial neural network model was developed to identify a writer from an unknown document and it achieved above 85% accuracy rate for each writer. In this article, writings of five Bangladeshi authors named Imon Jubayer (IJ), Humayun Ahmed (HA), Muhammed Zafar Iqbal (MZI), Kazi Nazrul Islam (KNI) and Hassan Mahbub (HM) are observed.