Academic journal article Issues in Informing Science & Information Technology

Automatically Generating Questions in Multiple Variables for Intelligent Tutoring

Academic journal article Issues in Informing Science & Information Technology

Automatically Generating Questions in Multiple Variables for Intelligent Tutoring

Article excerpt


Research on intelligent tutoring has produced many interesting results, for example Frasson, Gauthier, and Lesgold (1996), Larkin, Chabay, and Sheftic (1990), and Wenger (1987). Web-based online courses are also being developed at many educational institution and much research effort is focused on Web-based courses, for example, Boysen and Van Gorp (1997), Culwin (1998), and Wolz (1993). However, research on automatic question generation and difficulty analysis based on conceptual structures (Sowa, 1984) and ontological engineer (Heflin; 2001; "The Simple HTML", 2000) is still a weak link. Some researchers produced results (Soldatova & Mizoguchi, 2003; Kunichika, Katayama, Hirashima, & Takeuchi, 2003) using approaches not based on conceptual structures.

In a previous paper (Li & Sambasivam, 2003), we presented research results on automatic question generation and difficulty assessment for intelligent tutoring. That approach was successfully applied to the computer architecture course using a quantitative hierarchy. A knowledge structure, the concept graph, based on semantic networks, was used to automatically generate verbal, descriptive questions about computer architecture. That algorithm can only generate a question in a single variable. In addition, we also presented a quantitative method for assessing the difficulty of questions.

In this paper, the focus is on relaxing the restrictions of the previous algorithm. A new algorithm will allow the generation of a question in multiple variables. The knowledge structure for question generation is a Directed Acyclic Graph (DAG) concept graph.

This paper is organized as follows:

* This section presents an introduction to the subject of the paper. It also outlines the contributions of the paper.

* Section 2 is a review of our approach to automatic question generation. Concepts pertinent to the new algorithm are also defined in this section.

* Section 3 presents a new algorithm for automatic question generation in multiple variables.

* Applications and examples using the new algorithm will be presented in section 4.

* A summary with concluding remarks will be given in the last section.

Our approach needs well-formulated concept hierarchies. The development of a meaning hierarchy in a course may take years of experience and much time. Simply following a book or two will not be sufficient. We allow a combination of fixed object composition and dynamic multiple instantiations of a class of objects. Our approach is applicable to a wide range of subjects that do not require development of new data structures and algorithms. In other words, our approach, in its current stage, is not readily applicable to such courses as data structures and algorithm design.

The contribution of this paper is in its development of a new algorithm for automatic question generation on DAG based knowledge structure with multiple variables. This new algorithm is also guaranteed to generate solvable questions when the system of linear or non-linear equations are solvable.

A Review of Automatic Question Generation

A hierarchical functional concept graph is used in our teaching system. Our functional concept graph is an augmented version of conceptual graphs. A functional concept graph G consists of a set of nodes V and a set of directed edges E. The edges connect the nodes of V. In this paper, G is always a directed Acyclic Graph (DAG).

The functional concept graph is a hierarchical data structure: each node ni of the DAG is associated with a level [l.sub.i]. Node [n.sub.i] may have a set of incoming edges {[e.sub.i,1], [e.sub.i,2],... [e.sib.i,k}, connecting lower level nodes. A node without any incoming edge is a source node. A node is also associated with a value. The value vi of a node ni is computed by a function [v.sub.i] = [psi]([v.sub.i,1], [v. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.