Towards an authorship analysis of two religious documents

In this paper, we try to make an author identification of two ancient Arabic religious books dating from the 6th century: The holy Quran and the Hadith. The authorship identification process is achieved through four phases which are: documents collection, text preprocessing, features extraction and classification model building. Thus, two series of experiments are undergone and commented. The first experiment deals with authorship identification of the two books using a Manhattan centroid distance and SMO-SVM classifier. Whereas, in the second experiment a Hierarchical Clustering is employed to distinguish the different segments belonging to the two books. For that purpose, three types of original NLP features are combined before the classification process. The results show good authorship identification performances with an accuracy of 100%. In fact, all the results of this investigation correspond to a clear authorship distinction between the two religious books.