學習概念相關性之認知結構評量研究-以資料聚類與形式概念為分析工具

Concept-Relatedness-Based Cognitive Structure Assessment by Using Data Clustering and Formal Concept Analysis

陳嫻壎(Hsien-Hsun Chen)

指導教授 ： 賀嘉生

中原大學/工學院/資訊工程研究所/碩士(2005年)

https://doi.org/10.6840/cycu200500588

摘要

人類知識的認知結構為何一直是認知心理學家積極探討的，因為我們無法直接看到人類的心理結構，因此只能根據表徵(representation)的方法，將知識呈現出來，推論其內涵。教育學者應用網路模式表徵知識的目的在於，了解學生對於領域知識概念的了解程度，並希望藉由認知結構的測量與表徵結果，針對學生不理解的概念補強，讓學生能學得更好。
在眾多評量方法中，概念構圖法(concept mapping)與徑路搜尋法(Pathfinder)為兩種常用來診察學生認知結構的方法。概念構圖法是以概念圖表徵認知結構的方式，其包含概念(concept)與概念關係(concept relation)兩個元素，概念用來表示物件，概念關係表示概念間的關係。儘管如此，概念構圖法在實施上，需要對學生之構圖技巧加以訓練，才能建構出表達學生認知結構之概念圖。
而徑路搜尋法之測量分為三個步驟：第一，知識引出(knowledge elicitation) 收集學生對領域概念認知評定資料；第二，知識表徵(knowledge representation)步驟，以Pathfinder演算法將評定資料轉換成一網狀結構；第三，個別知識表徵的評鑑(evaluation of individual knowledge representation)，例如以數值代表認知結構。徑路搜尋法在知識引出步驟之實作上雖然簡單，但其知識表徵之網狀結構，常常只能以分數代表之，當需要了解學生概念之缺乏處時，則常常需要針對各節點逐一比對。
本研究的貢獻在於發展改善概念構圖法與徑路搜尋法之認知結構表徵方法，並套用結構取向之分析方法於認知結構評量上。除此之外，亦利用資料聚類與形式概念分析法分析出各類型學生發生錯誤之概念對為何？實驗的部分，以一份含有遺傳學12個概念之實際評比資料來分析，一共可分成34個聚類，除了有8類學生人數少於2人較無法討論之外，其餘28類各具特色，大部分學生都會在與配子或與DNA相關的概念對上發生錯誤，每類PFC指數與形式概念分析法分析之評鑑結果一致，除了從分數上來了解學生學習狀況，透過形式概念分析法更可分析出各種類型學生容易發生錯誤之概念對為何？藉此，可讓老師針對不同類型學生進行個別輔導，加強其所不理解之概念對。

關鍵字

認知結構 ； 聚類 ； 形式概念分析法 ； 概念圖 ； 概念相關性

並列摘要

Cognitive psychologists have devoted their efforts to understand the cognitive structure of humans. Because it is not able to be seen, knowledge representation is used to infer the intension of cognitive structures. In order to comprehend students’ understanding and discovery their weakness of domain knowledge, researchers and educators apply network model to represent mental knowledge of students.
Concept mapping and Pathfinder are two of common assessment methodologies. A concept map generated after concept mapping includes concepts and concept relations. Concepts are used to stand for objects while concept relations are used to represent the relationship between two concepts. Although concept map is effective in representing cognitive structure in a structural way, it needs training before students are able to construct the concept map which replies what they think in their minds.
Pathfinder on the other hand is a method which needs less effort in collecting students’ understanding of domain knowledge in knowledge elicitation step. After collecting students’ understanding, a network structure is produced according to Pathfinder algorithm. A numerical value is adopted as evaluation of each network structure. But it is not display more information than a value.
In this thesis, a methodology which integrates the advantages of concept mapping and Pathfinder is proposed. It converts the similarity rating data generated by knowledge elicitation step in Pathfinder into hierarchical concept map. In addition, formal concept analysis is applied to discover the characteristics of different type of students. The real data of twelve concepts of genetics are used in the experiments of this thesis. According to the results, the students can be divided into thirty-four clusters with eight outliers. The errors on the concept pair related to gamete or DNA are happened on most of the students. The experiment results in both of above methods are consistent. In conclusion, the method proposed here provides not only a way to represent cognitive structure but also more detail information of students’ cognitive structures for teachers to provide adaptive instruction.

並列關鍵字

concept map ； formal concept analysis ； clustering ； concept relatedness ； cognitive structure

參考文獻

[24] 宋德忠，林世華，陳淑芬與張國恩，知識結構的測量：徑路搜尋法與概念構圖法的比較，教育心理學報，30卷，頁123-142，1998。

[26] 江淑卿與郭生玉，不同學習過程的概念構圖策略對促進知識結構專家化與理解能力之效果研究，師大學報:教育類，頁1-16, 1997。

[44] 林原宏，知識結構分析-徑路搜尋、多向度量尺和集群分析的方法論探討，測驗統計年刊，卷4，頁47-69，1996。

[2] D.P. Ausubel, Educational Psychology: A Cognitive View, New York:Holt: Rinehart & Winston, 1968.

[9] B.J. King, "Step-wise clustering procedures," Journal of the American Statistical Association, vol. 69, pp. 86-101, 1967.