運用文件分群與概念關聯分析技術協助網誌瀏覽：任務導向評估方法

由知識管理的角度觀之，網誌(blog)提供一種新型態的知識交換模式，即網誌空間提供作者與讀者積極的互動轉換關係，促進知識的交流。由於大部份的網誌內容是偏於生活中的隨筆，其形式較無法像一般的網站，在設計之初就能規劃好所欲討論之內容及整體架構。本研究主要結合階層式分群法與概念關聯分析以協助網誌作者與讀者能夠正確的重新組織文章並進而有效的瀏覽文章。研究以使用者網誌的文章為例，設計考量使用者觀點之階層式文件分群方法將網誌之文章進行分群，以達成彈性分群之目的並由使用者角度重新組織網誌文章。研究並透過正規概念分析(formal concept analysis)方法建構與顯示各群組關鍵字詞之階層關係，主要目的為協助分群之命名並更有效的協助讀者進行資料的搜尋與瀏覽。研究並嘗試透過設計問卷與模擬使用者搜尋任務(task-oriented approach)以評估方法之有效性，實證結果顯示，結合分群與概念關聯呈現均有助網誌作者與讀者對網誌內容的理解，提升瀏覽效率與節省搜尋時間。

關鍵字

網誌；概念階層；正規概念分析；階層式文件分群；任務導向評估法

並列摘要

Blogs are a good online tool that encourages information exchange and knowledge sharing. However, blog users often face two challenges. First, blog contents are often categorized vaguely or inadequately by authors. Second, many popular blogs sort content by date. Consequently, when a blog user does not possess the right keywords for information retrieval, he or she must take time to browse the contents by chronological order until the relevant content is identified. This study addressed the problems by using the hierarchical agglomerative clustering and formal concept analysis methods to re-classify blog contents. To evaluate the effectiveness of the proposed automated text clustering solution, we further conducted user task-oriented evaluation. The results showed that the technique can help authors (bloggers) define new categories and refine existing categories. In addition, the concept hierarchy applied in each category helped the blog users to quickly discover the needed information.