從原型理論探討使用者的類別認知－以高中生網站分類歷程為例

A Cognitive Analysis of Categorization from the Perspective of Prototype Theory: Using High School Students Classifying Websites as An Example

毛宣蓉(Hsuan-Jung Mao)

指導教授 ： 藍文欽

國立臺灣大學/文學院/圖書資訊學研究所/碩士(2009年)

https://doi.org/10.6342/NTU.2009.01321

摘要

圖書資訊學的分類觀點朝使用者導向發展，許多研究認為需要多瞭解人們對於分類的認知。本研究試圖從原型理論的觀點，探討人們在分類過程中對類別認知的意象希冀有助圖書資訊分類者瞭解使用者對類別名稱、類別之間、類別與類別成員，以及類別成員之間關係的認知。
本研究以高中生為對象，探討其網站分類的歷程。以方便取樣與滾雪球取樣法，共訪談了16位高中生。研究者先篩選出22筆網站，請受訪者對這些網站進行關鍵字命名與分類，同時觀察受訪者實作的行為與反應。最後，以半結構性訪談法蒐集受訪者對網站分類的認知。
研究結果發現，16位受訪者所標示的網站關鍵字有分歧的現象。文獻中曾指出首位關鍵字有基本層次範疇的功能，但在本研究中，無法推論首位關鍵字具有基本層次範疇的功能。進一步觀察關鍵字類型，可歸納為主題型、網站特色、機構型與地區型。而受訪者給予網站關鍵字的目的有四，包括：檢索、對網站作專指性描述、概括性描述，以及作為網站與類別之間連結管道的特性。
受訪者對類別的認知，多以透過腦海中對類別的認識，直覺地分類網站。受訪者不會給予類別字典型的定義，同類別內的網站不是依據定義來聚集，而是彼此之間藉著相同部分，互相串連聚集。其中，網站內容具有「學」的意象，或是主題單一、內容的深入的網站，是受訪者覺得比較接近類別意涵的核心，能成為類別的代表實例。但網站即便是單一主題，若內容淺薄，也不會被認為是類別中可參照的對象。
對於「其他」一類，受訪者有不同的認知，歸納而言有三種類型：（1）認為「其他」是不屬於有名稱的類別；（2）「其他」是當網站與上述各類有關，但內容偏向生活、休閒性質，所歸的類別；（3）「其他」是聚集了相似網站，卻無法立即命名的類別。
類別與類別之間關係的類型，可歸納為類別的層級關係、互斥關係以及交叉關係。但是，人們對於類別與類別之間關係的認知，是較為模糊且相對的關係，與圖書分類的固定結構取向不同。
對於受訪者給予相同關鍵字的網站，可以歸納為網站之間有相同主題且面向相同，以及網站之間主題相同但不同面向的兩種關係。屬於前者的網站，它們之間的關係被認為比較相近，比較容易彼此聯想。少數受訪者會將相同主題不同面向的網站分在不同類，強調網站內容的不同。
在分類時，主題單一的網站，受訪者可以立即對應到網站所屬的類別，在分類上較為容易。對主題涉及層面比較廣的網站，受訪者會考量較多，有分類比較困難的現象發生。有些受訪者認為，如果網站涉及到多個類別的內容，就應該分到那些相關的類別中；但也有部分受訪者認為多元主題的網站仍應萃取出網站的中心主旨而分類，不會分到多個類別，以免無法表達網站的中心思想。
最後，提出對原型理論應用在圖書館分類與檢索取向的反思，對於未來類別認知的取向，也提出一些建議。

關鍵字

原型理論 ； 分類 ； 類別認知

並列摘要

In recent years, the user-oriented approach has gradually become the norm of classification research. Following the line, this Master’s thesis aims to explore how people classify Websites. From the perspective of Rosch’s Prototype Theory, this research attempts to conduct a cognitive analysis of categorization to seek a better understanding of the process and structure of categorization. It is hoped that the findings of this study will help the LIS field to device a more suitable classification scheme that meets users’ cognition.
The interviewees of this study were 16 high school students who were recruited to do a classifying task. Twenty-two Websites were selected in advance and the interviewees were asked to classify these Websites into five predefined categories. After the classifying task, a semi-structured interview was conducted with each interviewee to further investigate the thoughts underlining his/her classifying behavior.
It was found that the keywords used by these interviewees were dispersed. For most Websites, there was no consensus of the interviewees in choosing the keywords. In addition, these keywords could be grouped into the following four groups: Subject, Characteristics of Websites, Institutional factor, and Geographic area. The reasons why they chose these keywords could be grouped into four types as well: for retrieving, for generally describing a Website, for specifying the feature(s) of a Website, and to link a Website and its related category.
When classifying Websites into categories, most interviewees did not seem to have a clear definition of each category. They classified Websites based on their intuitions or an obscured image of each category. The Websites grouped in the same category were not assembled based on the definition of a category. Instead, there were brought together according to the likeness of each other and family resemblance.
Even though a Website is related to a certain topic, if it is too general or too superficial, the interviewees will not treat this Website as a representative or prototype of the category. They prefer to have a Website that provides more detailed or scholarly information of a topic. They would consider this kind of Website as a prototype of a category.
It is also interesting to note that these interviewees would use ‘Other’ category to denote the following situations: Websites not belonging to the five predefined categories, Websites that have similar topic/subject but different facet, and Websites with similar topic but the interviewees could not find an appropriate name to label it at the moment.
The relation between categories basically can be classified into two types: mutually exclusive and interrelated. However, for most categories, the line between two adjacent categories is not clear cut. The vagueness was mainly caused by the Websites containing multiple subjects. Some interviewees would classify this kind of Websites into multiple categories, therefore blurred the border line between or among categories.
In an interactive retrieval process, can we let people indicate a document or Website as a prototype of the information that he/she is looking for, and then retrieve more similar documents or Websites? Based on the findings, the investigator proposes that prototype can be considered as a factor for relevance feedback retrieval. The investigator also re-examines the appropriateness of applying Rosch’s Prototype Theory in LIS field and suggests some new directions for further studies.

並列關鍵字

prototype theory ； classification ； categorization

參考文獻

林金定、嚴嘉楓與陳美花（2005）。質性研究方法：訪談模式與實施步驟分析。身心障礙研究，3（2），頁122-136。

邱子恆（民91）。圖書資訊分類架構在組織與呈現知識上之應用。圖書資訊學刊，17，頁123-137。

萊科夫 (Lakoff, George)（1994）。女人,火與危險事物 : 範疇所揭示之心智的奧祕(Women, fire, and dangerous things : what categories reveal about the mind)（梁玉玲等譯）。臺北市：桂冠。（原作1987年出版）

楊敦淇（2002）。應用相關資訊回饋於貝氏混合式機率檢索模型。成功大學資訊工程研究所碩士論文。未出版，台南市。

Barreau, D. K. (1995). Context as a factor in personal information management systems. Journal of the American Society For Information Science, 46(5), 327-339.