Optimizing the Pre-Processing Phase of Automatic e-Document Classification

{orjeong, thlee, sglee} @ europa.snu.ac.kr Abstract. Electronic documents such as e-catalogs, e-mails, and Web documents have their own distinct characteristics that can be utilized in search and classification. They are structured, noisy, and, in some cases, related to each other. We analyze the characteristics of three major types of e-documents- e-catalogs, e-mails, and Web documents- and propose methods for optimizing automatic classification of such documents. Our Proposal is to improve the pre-processing phase of automatic e-document classification by considering the document-specific characteristics, while continuing to exploit the state of the art classification algorithms. We expect that the pre-processing techniques can be applied to other types of e-documents that exhibit similar characteristics