Genre, author and gender recognition in turkish texts using artificial immune systems

With the rapid growth of internet usage many documents arise in electronic form everyday in different categories. The increase in the number of documents arises the need for categorizing the likely documents in predefined groups. Documents are categorized to the predefined classses by a process called document classification. In this study, documents are classified in three main headlines which are according to their genre, author and author?s gender.Artificial Immune Systems are inspired from natural immune systems that are used in engineering applications to solve complex problems. Previously, they are used in pattern recognition, computational security, anomaly detection, optimization, machine learning, robotics control, tabulation, error detection and their branches, ecology, product systems, smart homes, adaptive noise neutralization, inductive problem solving, web server coordination, protein structure guessing and succesful results are gained but they are first used in document classification problem within this thesis study.In this study, 16 different feature vectors are constructed and tested on document genre detection, document author?s gender detection and authorship attribution with Artificial Immune Systems algorithms and mostly used classifiers in literature on document classification which are Naive Bayes, K-Nearest Neighborhood, Support Vector Machines and Random Forests.It is observed that dimension reduction techniques that are applied to the feature vectors increase the classification performance of classifiers.In the experiments it is seen that when we apply dimension reduction algorithm called YTU to suitable featurs vectors such as character n-grams, word roots and word stems and create new feature vectors, Artificial Immune Systems algorithms give successful results in document genre detection, document author?s gender detection and authorship attribution of Turkish documents and can be used in such systems.