本研究目的在於提出一種新的檢索策略,並與另外兩種檢索策略比較效果之差異。研究中建立一套智慧型資訊檢索系統,使其成為網際網路使用者之搜尋輔助代理人;首先,由使用者提出感興趣之樣本文件,透過基因演算法來組合關鍵查詢字串,如此於既有的文件字詞比對技術上做提高回召率的搜尋,再加上擷取部分段落文件作為使用者相關回饋單位,利用此延伸資訊藉以建立個別使用者之興趣檔,達到提高系統搜尋結果的精確度。 實驗之結果驗證本研究確實具有可行性,並推論出影響使用者興趣檔對檢索精確度的因素包含:興趣檔中關鍵字數量多寡、使用者若對該知識領域熟悉度高低、相關回饋進行之回合數。 This research aims to construct an intelligent agent to assist information retrieving on WWW. Then compare its effect with the other two information-retrieving strategies. Our intelligent agent requires the user to feed interested documents first to generate a user profile, then applies genetic algorithm to filter out the adequate query keywords string, and finally uses the user selected part of the retrieved documents as the relevance feedback to refine the user profile. Several experiments have been conducted. The results show that our strategy for information retrieving is better than the other two strategies.