Text Classification of news articles

Text classification is a very important area of research in machine learning. Naive Bayes is an efficient and popular classifier for Text Classification. But Naive Bayes is able to perform well when there are more different categories and gives poor result when the categories are similar. We compared the result of classifiers K-Nearest Neighbors and Naive Bayes on two sets of documents, one in which the documents are of mostly different categories and in other the documents are of somewhat similar categories. We also examine the performance of the classifiers by adding contextual and conceptual information to the data through addition of synonyms, titles of documents, phrases in place of words.