N-Gram Based Authorship Attribution in Urdu Poetry

Authorship attribution is an interesting problem in Computational Linguistics. Traditional author recognition systems for electronic text rely on techniques which train the system to the specific vocabulary and writing style of the writer and apply stochastic methods to judge a given text at byte, letter or word levels. In this paper we have developed a software system to apply one existing and one modified authorship attribution technique to Urdu poetry. This paper discusses the results of this exercise and also presents the issues and procedures related to the language model training and word level n-gram based recognition system development.