建構一個具有自動分類能力之電子郵件管理系統

Abstract

電子郵件對於現今的企業而言，提供了非常方便的溝通工具，而為了幫助企業解決電子郵件的管理問題，本論文提出並且實作了一個以SMTP為基礎的電子郵件管理系統，其功能包括流量控管、郵件備份、內容過濾、郵件分類、自動回覆和統計分析等。在我們的架構下，無論企業使用何種電子郵件伺服器，此郵件管理系統都能夠正常運作，不需依賴特定的郵件伺服器。而針對郵件分類的部份，我們以N-gram表示法為基礎，設計了一個中英文郵件皆可適用的分類演算法，不但能用在電子郵件分類上，亦可用於一般文件分類，經過測試後，證明此分類演算法準確率可達80%以上，並且具有學習機制，來修正和調整其分類結果。

The phenomenal growth of Internet accesses has made e-mail an essential communication tool of an enterprise. In this paper, we propose and implement an e-mail management system, which is based on SMTP protocol. The system is designed to resolve the content filtering, mail classification, automatic reply, analysis, auto feedback and statistics. Every company can utilize the proposed system without changing their mail server infrastructure. Moreover, we devise a N-gram based algorithm, which can be applied in both Chinese and English mails. This algorithm not only can be used in classifying e-mails, but also can be used in classifying other documents. The result of our simulation shows the accuracy of this algorithm can reach up to 80 percent and above. Also, the classification results can be revised or adjusted by the learning mechanism.