透過您的圖書館登入
IP:3.140.185.147
  • 學位論文

基於書目關係與使用者瀏覽路徑之網路書店連結推薦

Exploring Recommendations for Web Links of Online Bookstore: A Bibliographic Relationships and User Browsing Paths Approach

指導教授 : 陳光華

摘要


一個以使用者為中心的網站應該透過使用者熟悉的語言、文句和概念,使資訊能自然且符合邏輯順序地呈現,才能讓使用者滿意其瀏覽與檢索結果。網路書店乃充滿大量資訊的網站,書目資訊的呈現方式會影響使用者在瀏覽或檢索商品時的選擇,故其呈現資訊的方式需考量使用行為。若能透過瀏覽日誌檔分析瀏覽路徑,並得知使用者在瀏覽網路書店時會透過哪些書目關係來連結網頁,便能以此改善網路書店頁面上呈現的連結。 本研究的目的是探討使用者的瀏覽路徑、瀏覽樣式與瀏覽樣式所包含的書目關係,並以資料探勘的技術,經過清理、辨識區段及分割路徑來分析瀏覽日誌檔,接著從瀏覽路徑擷取使用者的瀏覽樣式,並計算各瀏覽樣式包含的書目關係數。最後分析總體與各類別書的書目關係特徵。本研究分析約 5 萬條瀏覽路徑,從中擷取約 150 萬種瀏覽樣式,平均每種向前瀏覽樣式有 1.17 個書目關係,向後瀏覽樣式則有 1.86 個。顯示使用者在得不到想要的資訊時會更依賴書目關係,故網站應提供符合讀者需求的書目關係連結。 研究結果發現,在總體書目關係中,使用者最常使用的書目關係依序是主題關係、責任者關係、經系統推薦而關聯、成員/群組關係和等同關係。本研究根據各類別書籍的書目關係特徵提出建議如下:1)將具有成員/群組關係的書目連結建置在文學小說頁面上;2)在各類書籍頁面上儘量建置本研究分析的五種書目關係連結;3)加強現有的成員/群組關係、主題關係及責任者關係連結;4)加強圖文漫畫與人文歷史類的頁面上能吸引使用者繼續瀏覽的要素;5)檢討現有的推薦準則與各類書籍的收錄標準。

並列摘要


A user-centered website should display information through languages, texts and concepts to which users are familiar and make information available in a natural and logical way as well. As a result, users may have better browsing and searching experiences. Online bookstores are full of lots of information related to books, and the ways they are displayed may influence on choices of users. Therefore, user-centered bibliographic information among books in web pages should be seriously considered by any online bookstores. If we know what bibliographic relationships (BR for short) of users’ browsing paths which were analyzed from user log files, we could modify the links on pages in online bookstores. The purpose of this study is to find out the user browsing paths, travel patterns, and BR of patterns. After cleaning user logs, partitioning user sessions, and identifying browsing paths from log files, we extracted the travel patterns from user browsing paths, and then figured out number of BR of travel patterns. Finally, we analyzed the features of BR by various categories. We analyzed about 50,000 paths, and extracted about 1.5 million travel patterns from paths. The average number of BR of each forward-pattern is 1.17, and 1.86 of backward-pattern. This means that if users can’t get information they want, they will rely more on BR. Therefore, website should provide BR links for users’ need. Research results showed that most frequent BR were “subject relationship”, “responsibility relationship”, “link by system’s suggestions relationship”, “membership/grouping relationship” and “equivalence relationships”. According to research results, a few suggestions are shown as follows: a) build “membership/ grouping relationship” link for web pages of fiction; b) build BR links analyzed in this study for web pages of each book as much as possible; c) strengthen current links of “membership/grouping relationship”, “subject relationship”, and “responsibility relationship”; d) strengthen attractive elements on web pages of graphic comic and history to attract users to continue browsing; e) review current guidelines for recommendation and survey the inclusion of various categories of books.

參考文獻


林盈霖(2009)。結合 HITS 之偏好瀏覽路徑探勘。未出版之碩士論文,靜宜大學資訊管理學系,臺中市。
蘇俊斌(2005)。應用網站探勘技術於網友瀏灠行為分析-以內容服務網站為例。未出版之碩士論文,國立臺灣大學資訊管理研究所,臺北市。
粘芸蓁(2009)。Amazon.com 圖書資料檢索結果中所呈現之書目連結關係探討。未出版之碩士論文,國立臺灣大學圖書資訊學研究所,臺北市。
陳垂呈、陳宗義(2009)。增進 Apriori 演算法探勘關聯規則。資訊科學應用期刊,5(2),71-58。
盧秀菊(2000)。圖書館目錄之書目關係。國家圖書館館刊,89(2),123-134。

延伸閱讀