以網絡書櫃資料建構讀者閱讀偏好多樣性之指標研究

Abstract

網路愛書人社群aNobii的興起，提供了讀者閱讀活動記錄的資料，讓我們得以研究讀者的閱讀偏好，並因此提升系統的書籍推薦表現。過去研究（Ross, 1999; Tang et al., 2012）藉由了解讀者的閱讀偏好結構，分析讀者的閱讀活動。其中，閱讀偏好結構包含偏好發展、閱讀偏好多樣性與閱讀涉入性。閱讀偏好多樣性在過去研究中顯示，對推薦書籍策略有影響，相較於其他兩個閱讀偏好結構的面向，相較容易量化，因此本研究選擇閱讀偏好多樣性為目標，希望藉由網路書櫃資料的建置，找出能呈現讀者閱讀偏好多樣性的指標。 閱讀偏好多樣性的計算，我們抽樣50個網路書櫃，並建立書籍的共現網絡，計算書籍間的相似性。接著，在5種相似性計算下，選擇3種分群方式，針對抽樣的書櫃做分群，分群數量則作為讀者的閱讀偏好多樣性，分群數量愈多，閱讀偏好愈多樣，反之亦然。由於過去研究顯示作者為重要的選書因素（Mikkonen & Vakkari, 2012; Tang et al., 2012），我們同時建立作者共現網絡，選擇書目計量學中計算研究跨領域多樣性的指標，以作者在個別書櫃中的種類與其所占比例做為計算多樣性的依據。此外，利用作者共現網絡計算作者間的相似性，將作者相似性加入多樣性指標的計算。為檢定個別書櫃分群數量與作者多樣性指標是否能呈現讀者的閱讀偏好，研究針對50個抽樣書櫃的擁有者進行問卷訪談，訪談結果則做為評量閱讀偏好多樣性的標竿。 研究結果顯示，Interminus相似性測量的分群結果與讀者自評多樣性呈顯著正相關，而作者多樣性指標結果則皆無顯著相關。Interminus相似性測量可大幅去除書籍僅被一個書櫃所擁有，而造成書籍相似性被扭曲的結果，而其餘4種相似性測量之個別書櫃分群結果，除與使用者自評多樣性皆無顯著相關，反之卻與個別書櫃之書籍總數呈高度正相關。綜上述，一般認為讀者閱讀偏好多樣性與閱讀書籍總數有關，然而，本研究閱讀偏好多樣性結果顯示，單從書籍總數無法判別讀者閱讀偏好多樣性，應進一步考量其閱讀書籍的相似性，才能準確地呈現讀者的閱讀偏好多樣性。

Usage data available through social media provides a great many opportunities to capture users’ preference. Using books saved in users’ online bookshelves, the study set out to explore social network analytical methods to capture the diversity of a reader’s reading interests. “Reading diversity” denotes how widely scattered one’s reading interests are. Drawing from data from aNobii, a social networking site for booklovers, users’ reading diversity was defined by the number of components created by the book co-ownership network of the books in their bookshelves. A total of 50 user’s bookshelf data were collected, resulting in a total of 21,199 distinctive books. They were also asked to fill out a questionnaire designed to elicit three dimensions of their preference: “reading diversity”, “preference insight” and “involvement.” Networks of the books were created where each node represented a book and the strength of their linkages were determined by five co-ownership based similarity measures: cosine, correlation coefficient, “normalized interaction”, and “intersection-minus-1.” The thresholds for the dichotomization of the five respective similarity measures were then determined by a level above which where the greatest percentage of the disappearance of the edges, which were then applied to the individual bookshelves so the number of the components in each bookshelf could be determined. Correlation analyses were then performed between the user’s self-assessed reading diversity and the number of the components in her/his bookshelf. One of the proposed similar measures, “intersection-minus-1” produced a clustering result that was significantly correlated with users’ self-assessed diversity. Furthermore, multiple repression analysis showed the proposed measure was able to provide explanatory power over and above mere counting the number of books in the bookshelf.