學生模組(Student model)是智慧型教學系統(ITS)中最關鍵的一個
模組，它的主要目的是擷取並記錄學生的知識狀態以達成調適性教學的目
的。本研究的主要目的是發展一套通用的學生模組化(student modeling)
的方法。適用的問題領域以程序性解題(procedural problem solving)領
域為主。 本研究有幾個特性：
(一)學生主動參與式：學生不再被動地被教、被考試，而是由學生主動教
系統解題的方法，系統從旁觀察學生的解題步驟，藉以了解學生的知識狀
態。
(二)具通用性：本研究所提出的方法有一部分具通用性，可適用在不同的
問題領域。
(三)可模擬性：本系統可自動產生一個解題器，這個解題器會模擬學生的
解題行為，產生跟學生一樣的解題過程與答案。
本研究所採用的方法：
(一)用context-free grammars來表示領域知識。領域知識是一個獨立的
模組，可以容易地修改而不須更改其他模組。
(二)用parse的方法來汲取學生的高階解題概念。
(三)汲取到的學生知識以context-free grammars來表示。
(四)解題器主要是由一個grammar expander構成，這個expander藉由展開
以grammar表示的學生知識來解題。
本研究已嘗試了三個問題領域，分別是解一元一次方程式、解二元一次聯
立方程組、及二進位加法。實作的結果顯示本研究所提的方法可適用於不
同領域的學生模組化工作。

Intelligent tutoring system(ITS) is an attempt to reform the traditinal computer-assisted instruction(CAI) and make it possess the intelligence of a human teacher. Student model is the key component to an ITS. A student model is responsible for acquiring and recording the student's knowledge state and makes it possible for adaptive teaching. The goal of this research is to develop a student modeling methodology for general purposes. The emphasis of this research is put on the acquisition of procudural problem solving knowledge. The characteristics of this research include: (1)student-initiative, (2) generality, (3) sumulatibility, (4) user friendliness. For these goals to be achieved, the following scheme is proposed in this research: (1) Domain knowledge is represented in context-free grammars. Domain knowledge is input to the system but not built-in. (2) High level problem solving concepts are abstracted through parsing. (3) The acquired student knowledge is represented in context-free grammars. (4) Problem solver is composed of a grammar expander and functions for primitive operations. The problem solver solves problems by expanding student knowledge in grammatical form. Three problem domains are tested in this research to assure the generality of the scheme. They are performing binary addition, solving linear equations in one variable, and solving system of linear equations in two variables. Results of the implementations show that the scheme can be applied in different problem domains.