The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

PhD Thesis #9319

Title:Recognising and Responding to English Article Using Errors; an Icall Based Approach
Date: 1993
Abstract:Artificial Intelligence techniques are increasingly being used to enhance the area of Computer-Aided Instruction. This thesis is concerned with the area of Computer-Aided Language Learning, a subset of Computer-Aided Instruction, and demonstrates how various Artificial Intelligence techniques can be incorporated into a language learning system to produce an intelligent educational tool. In this thesis, the focus is on the use of English articles, which is a subtle area of the English language with which even advanced students of English have difficulty. This thesis describes ArtCheck, an Intelligent Computer-Aided Language Learning (ICALL) system which detects, analyses and responds to English article usage errors. This system has three main features: it has knowledge of the article usage domain; it dynamically creates a model of the student; and it adapts to the individual student. The system's knowledge of the domain consists of a set of article usage rules which reflect standard teaching practice. The information necessary to apply the rules is extracted at the natural language processing stage, and includes structural and contextual information. The system models the state of the student's knowledge at all times, in order to give informative explanations to the student about any errors which are made. It is able to generate mal-rules which account for consistent errors made by the student, using version spaces and the candidate elimination algorithm. the student model can be described as dynamic because the generation of mal-rules can create new parts of the student model, in response to student behaviour, which are not predetermined by the system designer. the system responds to individual students by giving explanations of errors which are tailored to the student's level of ability and preferred learning style. The type of explanation given is also dependent on the system's assessment of the source of error, and reflects any mal-rule which the s

[Search These Pages] [DAI Home Page] [Comment]