Motivated by applying Text Categorization to sorting Web search results, this paper describes an extensive experimental study of the impact of bag-of-words document representations on the performance of five major classifiers -Naive Bayes, SVM, Voted Perceptron, kNN and C4.5. The texts represent short Web-page descriptions from the dmoz Open Directory Web-page ontology. Different transformations of input data: stemming, normalization, logtf and idf, together with dimensionality reduction, are found to have a statistically significant improving or degrading effect on classification performance measured by classical metrics - accuracy, precision, recall, F_1 and F_2. The emphasis of the study is not on determining the best document representation which corresponds to each classifier, but rather on describing the effects of every individual transformation on classification, together with their mutual relationships.
展开▼