Comparison of the Performance of Genre Classifiers Trained by Different Machine Learning Algorithms

Modern search engines aim at classifying web pages not only according to topics, but also according to genres. This paper presents the results of an attempt to train a genre classifier. We present features extracted from a 20-genre corpus used for training the genre classifiers and the results of using different machine learning (ML) algorithms in the process of learning. Success of the genre classifiers was measured by accuracy, precision, recall and F-measure. Accuracy did not turn out to be a good indicator of classifier success. In the case of other measures the results show that different algorithms should be used for training purposes depending on whether the user wishes to obtain high precision or high recall.