Predicting the Gender of the Kurdish Writers in Facebook
Peshawa J. Muhammad Ali
Department of Engineering Software, Koya University
Received : 5/7/2013, Accepted : 17/11/2013
DOI Link: https://doi.org/10.17656/sjes.10010
Facebook is one of the social networks which have lots of users among Kurdish people. Although there are no formal or published statistics about the number of the Facebook users, in the last few years Facebook was the most used website among Kurdish society. This swift development of the Kurdish society towards Facebook imposes new challenges that need to be addressed. For example, a poem or an article published on Facebook possesses properties such as author name, gender, age, and nationality among others. In this paper the gender of Kurdish authors in Facebook determined by using a feed-forward artificial neural network model. 120 Facebook Kurdish written posts were used for learning the model designed to determine the gender of Kurdish writers in Facebook. The posts were taken from Facebook pages of different persons with different backgrounds. Twenty eight text features were extracted from each post; these features were distinct in discriminating between genders. The feed-forward back-propagation artificial neural network with three layers (28 nodes, 14 nodes, 1 node) is used as a classification technique. The accuracy ratio which based on the ten-fold technique (taking the average ratio among ten trials) obtained was 77.5 %. This proposed idea of this paper is important for detecting the real gender of Facebook page owners.
Facebook, neural networks, text mining, gender identification.