Performances Evaluation of Automatic Authorship Attribution on Ancient Arabic Documents

Authorship Attribution is a research area concerned with the automatic classification of text documents based on their authors. The principal objective of this investigation area is to discover the author of a given text is. This task becomes very tough as far as old text documents are concerned. In this paper, we attempted to broach the Authorship Attribution problem as it applies to old text documents. For this purpose, several experiments are conducted and their results are commented. In order to validate the performances of our system, we constructed a special dataset that we called ''10AAP '' (10 Ancient Arabic Philosophers), by quoting texts from the works of 10 ancient Arabic philosophers, where the topic of the different texts is the same. Moreover, the genre of the authors is also the same.