數學詳解影音自動化生成系統之設計原則與架構

Abstract

隨著數位科技的進步，數位學習（e-Learning）逐漸成為當今熱門的學習模式。然而，傳統影音教材的製作歷程卻無自動化設計，導致影音教材的重用性低落、製作時間成本高，無法因應學習材料的多元性需求。本篇研究回顧數位學習和數位影音教材的發展脈絡，並結合多媒體學習理論、科技批判的角度，彙整出未來數位影音教材製程應具備的四項特性：「可重用性」、「可互動性」、「可調整性」和「學習效果還原」。

作為理論的實踐，本篇研究提出數學詳解影音自動化生成系統（explanatory math animation auto-generating system, EMA system），此系統能夠自動分析解題過程中會用到的數值資料，並能將該資料與 Manim 動畫引擎、 Microsoft Azure 語音生成服務結合，讓一部原長 118 秒的數學詳解影音，平均僅需24.428 秒即可製作完成。此技術相較於傳統影音教材的製程，除了大幅縮減影音教材的製作時間，也能解決現今影音教材的多元性需求。

With the advancement of digital technology, e-Learning has gradually become a popular learning method. However, the traditional process of producing instruction videos lacks automated design, resulting in low reusability, high production time costs, and the inability to meet the diverse needs of learning materials.This thesis reviews the development of the production process of multimedia learning material. This thesis also combines the perspectives of multimedia learning theory and technological criticism to summarize four characteristics that future video learning materials auto-generating systems should have: "reusability," "interactivity," "adjustability," and "learning effect restoration."

Besides, this thesis proposes the Explanatory Math Animation Auto-Generating System (EMA System). EMA Systemcan automatically analyze the numerical data used in the problem-solving process and combine the data with the animation engine - Manim and the speech generation service. In EMA system, the production time of a 118-second explanatory math animation is an average of only 24.428 seconds. This technology can significantly reduce the production time of instruction videoscompared to traditional processes.