%0 Journal Article
%T Detecting the Authors of Texts by Neural Network Committee Machines
%A Alen Savatic
%A Amir Jamak
%A Mehmet Can
%J Southeast Europe Journal of Soft Computing
%D 2012
%I 
%X This paper proposes a means of using a boosting by filteringalgorithm in artificial neural networks to identify the author of atext. This approach involves filtering the training examples bydifferent versions of a weak learning algorithm. It assures theavailability of a large source of examples, with the examples beingeither discarded or kept during training. An advantage of thisapproach is that it allows for a small memory requirement. Once thenetwork has been trained, its hidden layer activations are recordedas a representation of the selected lexical descriptors of an author.This stored information can then be used to identify the texts writtenby the same author. Texts studied are literary works of two Bosnianwriters, Ivo Andri  (1892-1975) and M. Me a Selimovi  (1910-1982).The data collected by counting syntactic characteristics in 1466paragraphs of "na drini  upria" by Ivo Andri , and "dervi  i smirt" byM. Me a Selimovi  each.
%K machine learning
%K author identification
%K artificial neural networks
%U http://www.scjournal.com.ba/index.php/scjournal/article/view/13/12