Security Level Classification of Confidential Documents Written in Turkish

This article introduces a security level classification methodology of confidential documents written in Turkish language. Internal documents of TUBITAK UEKAE, holding various security levels (unclassified-restrictedsecret) were classified within a methodology using Support Vector Machines (SVM’s) [1] and naïve bayes classifiers [3][9]. To represent term-document relations a recommended metric “TF-IDF" [2] was chosen to construct a weight matrix. Turkic languages provide a very difficult natural language processing problem in comparison with English: “Stemming”. A Turkish stemming tool "zemberek" was used to find out the features without suffix. At the end of the article some experimental results and success metrics are projected.