基於本體論數位學習內容擷取之研究

Abstract

內容擷取對於數位學習應用發展是非常重要的一環。然而，一般的資訊檢索技術著重於關鍵詞層面的處理，對於詮釋資料、文件結構與概念層面著墨較少。事實上，數位內容的擷取若能支援教育理論的應用，例如：教學策略、學習風格等，對於教學成效的提升將更有助益。有鑑於此，本論文採取較高層次的觀點來處理數位學習內容。希望能針對新興的數位學習環境(格網環境、環境感知學習、同儕網路)，提出有效的內容擷取方法。近年來，本體論已成為探討知識領域概念化的重要模型。因此，本文提出基於本體論的模型與方法，從上層概念的觀點來組織與檢索內容，然後將此模型實現在三種新興的數位學習環境。首先，我們針對集中式格網環境提出「由下而上」的方式，為大量且分散各地的內容建立索引。其次，我們為環境感知學習設計了一個支援教學策略的查詢擴充法。接下來，我們探討同儕網路上的等待回覆時間設定問題。此外，我們也說明如何建置本文所需的本體論，並利用「庶民分類法」來調整本體論與註記內容。綜上所述，我們將這些想法應用在教材設計，提出了一個「維基式」方法。我們實作了這些方法，並針對國小師生進行相關實驗。評估結果顯示本文提出的方法可以達到有效且快速的內容擷取。問卷調查結果也顯示使用者認為這套工具有助於取得適合的學習內容。

Content retrieval is important for the development of e-learning applications. However, existing information retrieval techniques focus on keyword-level processing, rarely considering metadata, structural information and concepts of contents. In fact, content retrieval can increase learning performance by supporting pedagogical theories and models, such as instructional strategies, learning styles, etc. Therefore, this dissertation addresses learning content from a high-level viewpoint, and proposes effective and an efficient content retrieval approach for emerging e-learning environments, such as grid environments, context-aware learning and peer-to-peer networks. In recent years, ontology has been widely investigated to model and conceptualize domain knowledge. Hence, an ontology-based model is proposed to represent, organize and retrieve content from a high-level viewpoint. Then, the proposed approach is realized in three emerging e-learning environments. First, a bottom-up method is proposed for centralized grids to create indices of contents in widely spread repositories. Next, a knowledge-based query expansion method is designed to support instructional strategies for context-aware ubiquitous learning. Then, the due-time setting problem for retrieval in P2P networks is addressed and solved by an interactive method. In addition, the ontology construction is described. A folksonomy-based method is proposed to refine the ontology and annotate the learning content. Next, the above-mentioned ideas are applied to teaching-material design by a Wiki-based method. The aforementioned methods are implemented, and experiments are conducted in an elementary school. The results of evaluation show that the proposed method is efficient and effective. Surveys of user opinions also show that the proposed approach can assist learners to retrieve appropriate learning content.