通俗分類標記之自動組織

Abstract

網際網路上的資訊量成長得相當快，藉由網路上社群的力量，資訊的分類能由大眾來做決定，也就是所謂的通俗分類 (folksonomy)，而這些通俗分類中的標籤仍然可以有進一步的組織。本論文中使用了兩種方法將 del.icio.us （一個協同標記系統的網站，能讓使用者儲存並註記自己的網路書籤）中，由搜尋所產生的結果中的共通標籤加以重新整理、分群。 我們將自動產生的分群結果與熟悉相關主題使用者手動分群的結果相比，可得知這兩種自動產生的方法與使用者產生結果的相似度相距不遠，而不同使用者產生的群聚結果也大不相同。我們亦施行了理解度測試，以了解使用者是否能從產生的群聚結果中獲得助益。我們認為，組織分群後的標籤的確能讓使用者更快熟悉其搜尋的概念，以及幫助其導覽。

Information grows fast in volume on the Internet. By the power of the Web communities, clasffication of information is done by the public -- the 'folksonomy'. However, we still want to further utilize those collaborative generated tags in the folksonomy system. In this thesis, we survey two approaches to organizing and clustering the common tags of the search results from a collaborative tagging system 'del.icio.us' which let users store and annotate their bookmarks. We compare the automatically generated clustering results with those by humans who are familiar with the topics. We observe that the results of the two approaches we survey have close similarities with those by the people. And even people can not agree on the clustering. We then conduct a comprehension test to see if people really benefit from the results. We believe that the organized tags really help users in framing their concepts and navigation.