Language identification (LI) is a phase of natural language processing. Although LI is formerly studied, there is still much work to do for better performance. The purpose of this study is to present low dimensional feature set which is built from letters and diacritics and suitable classification algorithm (C-SVC, MLP or LDA) with it for high performance. In addition, a weight factor has been integrated to language identification system for increasing the performance. Experiments have been done on ECI corpus. Weight factor has increased the classification accuracies. The most accurate and the fastest method is C-SVC for our feature set. (C) 2011 Published by Elsevier Ltd.