近年來隨著線上社群平台的興起,使用者已不再只是單純被動的資訊接收者,而可以自由分享與提供資訊,進而產生互動,漸漸形成社群網路。 目前,使用者對於社群網路相關但不甚了解的資訊,通常會利用網路上的搜尋引擎系統進行關鍵字搜尋,但因為資訊量過於龐大,使用者可能需要一些延伸建議進行適當的修改原始查詢,以縮小搜尋結果範圍,快速得到正確結果。然而,目前的查詢延伸的技術大多是基於整個網路的使用者資訊得出,卻忽略了社群網路的使用者關係,也忽略了社群內使用者群體智慧的力量,因此無法幫助使用者探索社群內其他使用者所擁有的潛藏資訊。因此,本研究提出了針對使用社群書籤服務之使用者,利用其使用者間的社群互動、收藏等資訊,給予查詢延伸建議的方法,希望幫助使用者進行網路搜尋。 本研究收集了delicious網站使用者的收藏與其自行命名的標籤等資訊做為研究資料來源。我們利用使用者間互動的關係建立社群網路、以收藏資訊建立超空間語言相似度分析(Hyperspace Analogue to Language, HAL)空間,進行查詢延伸字的計算,再利用社群網路關係資訊,以多個面向給予查詢延伸不同權值進行分析,找出最適當的結果進行查詢優化。Recently, with the rising of on-line community platform, users are not only the information-receiver anymore, but can share and provide information easily. They can interact with others, and it forms a social network gradually then. Users usually use web search engine to search the information which about the social network they don’t know, but the information is too much, so they might need some query expansion suggestion to revise the initial query, then reduce the range of searching, and get right result quickly. However, most of the current query expansion techniques are based on all the user information in the internet. It ignores the users’ relationships and the power of collective intelligence in the social network, so we could not help users to explore the other users’ hidden information. This thesis proposes the way to help the social bookmarking services users to improve the result of web search by using the interaction and collection information. We collect information of users on the website named ”delicious.” We form the social network based on users’ interaction, and build up the HAL(Hyperspace Analogue to Language) space to calculate the expansion words, then use the user relation information of the social network, analyzing by different dimensions to identify the most appropriate result for query refinement.