智慧型電腦輔助學習環境中超媒體學習型態之分析

Abstract

在超媒體上的需求式教學已經成為透過網路的遠距教學的重點之一。在這 篇論文、我們提出一個量化的模式去做在超媒體上的學習分析。雖然從前 就有人對超媒體上的學習分析相當的重視，但是僅有少數的研究者對學習 模式用客觀的學習時的數量化的資料做分析。大多數現在的分析方法都依 靠人類專家依一些合適的分群項目將學習模式分成數類。這些方法都太主 觀。我們在這篇論文裡將呈現用現代人工智慧科技去分析學生在超媒體環 境中的學習活動的資料的優點。在這篇論文中，我們從知識結構與教育超 媒體的架構的關係開始討論，然後討論在超媒體瀏覽的非線性問題。再來 我們提出幾種方法來找尋學生在使用教育超媒體時的學習模式。我們使用 最大相同子序列 (LCS) 計算兩個學習序列 (路徑) 的相似程度，使用模 糊分群法配合 LCS 將學生的學習路徑做分群。我們還定義一種方法去計 算學生的學習路徑的非線性程度 (Hyper Degree)。我們並使用模糊類神 經網路去找尋學習活動與學生背景的關係。在上面所提方法我們都將之用 在「空中英語教室」這個多媒體教材上。並且我們還將計算非線性程度的 方法用在分析真正超媒體上的瀏覽路徑。 Hypermedia course-on-demand has become a focus of distance education through computer networks. In this thesis we propose a quantitative model for hypermedia browsing pattern analysis. Although the importance of navigation behavior analysis has long been addressed in the past, only a few researchers have discussed how to classify patterns based on objective, quantitative data recorded during a learning session. Most existing methods rely on human expertise to divide the patterns into several categories and then characterize these categories with proper terms. Conclusions drawn this way are largely subjective. We show in this thesis how to take advantage of the modern technology of computational intelligence to analyze the information about student learning activities in a hypermedia environment. We first define measurement indices for a hypermedia tutoring system. Then we introduce a quantitative approach to determining the similarity between navigation patterns. We use the longest common subsequence of two browsing paths to indicate their partial resemblance, which, together with other metric measures, provides a sound basis for similarity analysis. Models for categorization, including fuzzy clustering and neuro-fuzzy association, are then described to complete this quantitative model. We believe that this model, working together with a network recording module, suggests an approach to student modeling in hypermedia-based distance education.