電子化政府是政府部門於便捷的網路環境上提供便民服務，讓民眾可以不用出門便 可處理各項業務。而為了提供民眾申訴或表達意見的管道，在政府部門網站內提供「首 長信箱」功能。為表示對於民眾意見的重視，因此必須快速且正確地將民眾意見分案至 相關單位進行處理及回應。本研究透過訪談方式，發現在這些分案專家的腦中似乎有張 各部門工作職掌的知識地圖一般，可以快速且準確的做好分案工作。過去關於文件分案 的研究，都忽略這一塊，而直接用文字探勘的技術來做分類。基於上述緣由，本研究嘗 試提出二階段的分案處理模式建立文件分案自動化的機制，減少人工作業的流程及成 本，提升政府的服務效率。本文所提之二階段分案處理模式，首先運用訓練資料及文字 探勘技術來建立知識地圖，接著透過新進文件與知識地圖的比對進行文件分案的預測。 經實驗證明，本文所提出之分案模式可以準確且有效率的達致預期的目標。

E-government means using information and communications technologies to improve the transparency, efficiency, and effectiveness of public institutions. The government can provide many convenient online services to serve the civil well. Especially, many local governments in Taiwan have used “Mayor’s Email-box” to handle citizen complaints or opinions. Quick responses to those complaints or opinions are needed in e-government services. Therefore, to classify the citizens’ opinions quickly and correctly has become an import work for the government. Through interview with the domain experts, we found that “knowledge map” can improve the efficiency of classification system which was ignored before. This study proposes a framework which contains two phase works. First, the knowledge map should be built up based on the association rules discovered from the training data. Next, the unclassified documents will be categorized according to the knowledge map. To verify the proposed framework efficiency and accuracy, extensive experiments are conducted using real data sets. The experimental results indicate that the proposed method is computationally efficient, and can effectively categorize documents.