A comment text in large enough quantities can be processed into something useful for the development of an e - commerce. Text commentary structurally complex and incomplete, meaning vague and not standardized, and different languages plus a translation is not accurate. In this study it was found that a comment text that many causes of customer service division of labor is less than optimal because the staff were inconsistent. So the need for text mining analysis is needed in dealing with the comments of large amounts of unstructured and such. One of the most important activities in text mining is classification or categorization of text. Text classification method used is Naive Bayes classifier, with data processing and data testing training. Comments will be classified by categories comments are often sent by customers Noonaku Signature through the site. Comments will be categorized into three classes, among other bookings (orders), production (product), and payment (payment). This research will be generated by means of detection accuracy rate of manual and automatic detection and the result indicating that the study provides good results in the classification.