İkili Arka Plan Destekli Yazar - Bağımsız Yazarlık Doğrulama Sistemi
xmlui.mirage2.itemSummaryView.MetaDataShow full item record
Authorship verification is a decision problem that tries to find out whether the text of an unknown author belongs to a suspicious author. The problem is the fundamental one of the studies in authorship analysis that has been interested in many years. The challenges of the authorship verification problems have arisen from the small number of known documents of the suspicious author. In the literature, author-dependent style comparison has generally presented as a solution to the authorship verification problem, which is a current and interesting subject. However, as the amount of known text of the suspicious author decreases, it becomes harder to extract accurate style. In this thesis, studies have been carried out on how authorship verification should be done even when the amount of known texts from the suspicious author is very low. Within the scope of this thesis, a binary background assisted author-independent system is proposed for the solution of the situation we handled. The proposed system has more generalization ability than the current solutions. Besides, the proposed system is language independent. The system, in which different languages can be easily integrated, has been tested within the scope of the thesis with Turkish and English languages, and the most accurate authorship verification result has been obtained in a public English dataset used in the field of authorship analysis. In this thesis, also, a Turkish Blog dataset has been collected and shared with the researchers in order to increase the Turkish-based authorship analysis studies. The proposed system, which also produced successful results when using the Turkish Blog dataset, has produced results that will be useful even in determining the high discriminative features with the priority of the Turkish language.