Text classification is the most important research issues in the field of data mining. The main idea of using the stemming technique is to reduce the number of features that can be extracted from the document. Furthermore, the stemming aims to enhance the accuracy of the classifier. This paper aims to study the effectiveness of using stemming techniques. The paper will use two popular word extractions: Khoja and Light stemmers. The results will compare with the result of classification without using the technique of word extraction. In the experiment, the Sequential Minimal Optimization (SMO), Naive Bayesian (NB) J48 and K-nearest neighbors (KNN) were used to build the training models and test the data. By implement the two approaches of word extraction and measured the accuracy of them by precision, recall and f-measure, the results show that the Light stemmers outperforms the Khoja stemmer. Furthermore, the results were comparing with the results of classification without using stemming technique.