Web 2.0概念的圖書館個人化推薦系統

Abstract

Web 2.0的精神是想透過社群的集體力量，創造、分享並評論屬於使用者自身或他人觀點的內容。而自2002年以後，從Wikipedia可以證明此種以群眾意見為基礎的內容創造及評論模式，不論在客觀數據及社會觀感中均具有可信的份量。

而在Web 2.0之前，網路商店提供的推薦清單，個別使用者並無法回饋有關推薦精確度的訊息，而其他使用者也無法自他人的意見回饋中獲益。

因此本論文將Web 2.0的精神與個人化推薦系統相結合，並應用在圖書館推薦系統中。本論文使用資料探勘 (Data Mining) 的協力式過濾 (Collaborative Filtering) 得出個別讀者的推薦清單，再經由讀者們對書籍難易度的評價、與個別讀者設定難易度的等級，過濾出難易適中的推薦書籍；同時經由讀者們對書籍tagging等Web 2.0的活動，重新對館藏進行分類，使得圖書館的藏書以一種更貼近當代讀者的面目呈現，以消除讀者對圖書館的隔閡。

本論文希望由量變產生質變，透過社群參與難易度的評價。讓讀者自行決定書籍的適當閱讀順序與分類，讀者將透過更親切、容易的方式找書，同時也讓前人的閱讀經驗得以留存，幫助後進者的求知之路。

True to the Web 2.0 spirit of creating, sharing and tagging by open-communities, the content of websites are no longer provided by site owners but users. After 2000, Wikipedia, as one of the paradigms of Web 2.0 websites, proved that this kind of running model which made of people, tagging and review by people has earned trustworthy reputation in objective data and general impression.

Before Web 2.0 era, users could not response their feedback to recommendation list of online stores, with the result that stores could not improve the system by collecting feedback.

In order to solve the problem and offer an adaptive recommendation system that automatically adjusts recommendation result to users’ preference by collecting response of users, our research combined Web 2.0 features with personal recommendation system and put in use in library. First of all, our research applies Collaborative Filtering, one of solutions of Data Mining, to obtain individual recommendation list. Secondly, system filters out unsuitable results depends on personal rating records, and in proportion to overall rating by all of the users. Therefore, the final recommendation list should be more accommodate to each user. Furthermore, users’ tagging would also reconstruct library catalog, which could break down the barrier between library and readers.

Our research expects of quantitative and qualitative change phenomenon by encouraging community participation. Hence, by archiving and analyzing forerunners’ rating, review and tagging records, following readers would be easier to find right books quickly.