協同研究系統之標註與文獻推薦技術

Abstract

近年來，線上協同研究系統(Collaborative Research Systems)變得愈來愈熱門，因為它提供了一個良好的研究環境讓使用者可以在線上互相合作、管理並且分享彼此的文獻。傳統上使用者在分類自己的文獻時，常常因為不知道要如何明確分類而苦惱。而比較明確的分類法，像是ACM分類法(ACM Computation Classification)，對於一般使用者更是難懂而難以使用。而大多數的協同研究系統之所以受歡迎的一項特色乃是它提供使用者利用標註(Tag)管理文獻的功能，因此使用者可以簡單自由地自行定義自己喜歡的關鍵字來管理組織自己的文獻。然而，使用標註來管理文獻雖然容易而且廣受歡迎，但是若使用者遇到自己不熟析的文獻時，使用者可能沒有辦法對此文獻下比較明確的標註。此外，當協同研究系統中的文獻累積到很巨大的數量時，使用者將會花更多的時間在尋找他們有興趣的文獻。 在此論文中，我們提出了一套標註推薦的機制來針對文獻推薦標註，用以協助使用者更能容易的管理文獻。此外，我們提出了兩套文獻推薦技術，分別是Tag Match和Tag-based CF，用以幫使用者找出他們有興趣閱讀的文獻。最後，我們利用從CiteULike擷取的資料來進行實驗，實驗結果說明我們所提出的標註與文獻推薦機制是簡單而且可行的。

Collaborative research systems recently have grown in popularity on the web that allow users to collaborate on organizing and sharing bibliography online. Since it is difficult for users to categorize bibliography, especially in some professional way, e.g. the ACM computation classification, nowadays the most important feature provided by collaborative research systems is that they support tagging mechanisms so that users can easily and freely define their favorite keywords to organize bibliography. Although tagging is a simple and popular process for organizing bibliography, sometimes users still have difficulties to tag a bibliography if they are not familiar with it. Besides, if the bibliography collections in a collaborative research system have grown to a large amount, users will spend more and more time on finding bibliography in which they are interested. In the thesis, we proposed a tag recommendation mechanism which can suggests tags to help users organize their bibliography, besides, we proposed two bibliography recommendation mechanisms, Tag Match and Tag-based CF, to recommend users bibliography they may be interested in. We conduct an experiment using the dataset clawed from CiteULike to show that the tag and bibliography recommendation mechanisms developed in this work are simple and feasible.