Skip to main content

REPAIRING ERRORS IN COMPUTER AIDED LANGUAGE LEARNING

Objective

RECALL is working towards more adequate and user-oriented software for computer-aided language learning by investigating intelligent feedback facilities as an enhancement to an existing commercial prototype. Prospective end-users include tutors and students, educational CD-ROM suppliers and corporate training departments. The system will enable the proficiency and strategies of individual learners to be dynamically profiled. It will detect errors in their input and offer feedback helping them to understand and avoid recurrence of their mistakes.
RECALL has investigated the requirement to provide error correction as an enhancement to a prototype CALL package. The facility detects errors made by students and provides them with feedback in a way which helps them to understand the nature of their mistake and to learn from it. Such a facility is designed to offset the lack of flexibility and feedback in current systems.
The project has enabled progress towards more adequate and user-oriented CALL software by taking account of individual learner strengths and weaknesses, levels of proficiency, and preferred learning strategies. It will serve as a pilot study for future activities in the area of intelligent CALL applications.
The package incorporates a module that provides detailed linguistic analysis of the learner's response to exercises (a diagnosis component), a module that creates a dynamic model of the learner, representing his learning history (a learner model), and a module that controls the system's reactions to the learner's input and the structure of materials offered (a tutor model).
Progress
In the first year, work focused on a market survey and user requirements. A preliminary market report, based on the results of primary and secondary research was then finalised and distributed. The user requirements report is based around both structural and functionality factors in both the existing and the enhanced package. The results of the market and user studies led to changes in the design of the intelligent error correction component and have influenced future developments for CALL applications within the consortium.
The consortium set up two user groups, one for English as Foreign Language, the other for German. User meetings have been held for both groups and a report on these meetings profiles the members. A project help desk has been set up to maximise user participation and build on existing work to secure active and varied user group commitment and participation.
User Requirements
The user study aims at enhancing the functionality, usability and effectiveness of CALL systems in general, and the CASTLE demonstrator system in particular. Functionality was assessed using observation, questionnaires and interviews. The results were used to determine the structure of a CALL system and to design a functionality checklist to be used in quality assurance procedures and processes within system development.
Analysis of user requirements has involved two different classes of user:
- language learners or end users
- intermediate users such as CALL product suppliers, authors, tutors, publishers and language training centres.
Members of the latter class include many high profile European names in the business world.
Different techniques based on ISO standards were used to measure two prototype system versions. Comments and problems associated with each prototype were documented in a Usability Study report, which in turn was fed back in the software development process.
Further work has been completed on comparing learner reaction to 'traditional' courseware and to courseware with additional access to the CASTLE system. Protocol data from Castle has been used to obtain information about error situations and their contexts. Work to gather didactic information on which to base additional teaching strategies for CALL systems is on-going.
Requirements data was collected from 319 subjects.
The main important findings are:
- social interaction and business scores high for topics, holiday and travel are less important
- choice of style ranks less important than good coverage of vocabulary and grammar
- learning context should include home, work or language training organisation
- comprehensive coverage of learner grades is needed
- systems should not be used without guidance and competitive situations should be avoided
- ways of integrating the tutor's own material into the system is needed
- enhanced test procedures covering exam simulations are not necessary
- additional features such as word processing or interfaces to other packages would be useful.
Market Survey
The personal market sector for CALL products is experiencing the strongest growth, estimated at 50% per annum. The main influence in this sector is the increasing range of 'edutainment' CD-ROMs. Preferences are for attractively packaged products which are fun to use.
In the vocational training sector, large enterprises are the main users of multimedia educational tools, mainly as a result of bigger purchasing power compared to SMEs. High levels of computer literacy among executives and a positive attitude to learning technology are set to increase the size of the corporate market. Trends are towards home learning rather than training in company time which has given rise to increased interest in distance learning.
In the basic language training sector, growth has been inhibited by the lack of adequate basic IT infrastructures. This is changing and some countries have already launched initiatives to stimulate the use of CALL systems in this sector. Some growth is this sector is also coming from factors associated with the market situation in the personal sector. An emerging sub-segment in this sector is the primary school market.
Growth in Internet use has also led to more commercial initiatives in training and education. Better, faster and cheaper options for communications facilities will further enhance growth in the distance learning market.
Artificial intelligence technologies have led to improvements in CALL system features and functions in what has been termed ICALL (Intelligent CALL) or ITS (Intelligent Tutoring Systems). The development of ICALL reflects the shift from products which apply technology to a specific point in the learning process to products which act as a progressive guidance system that takes learners from beginners, through intermediate and on to advanced levels.
Components are now being linked together via hypermedia in order to provide a greater degree of learner independence and a more complete language learning experience in ICALL systems. An adequate error feedback component, as proposed by RECALL, is widely considered to be the missing link in the provision of 'learning' material.
System Architecture
The system architecture integrates the following components in the CASTLE CALL system:
- a Diagnosis component that provides detailed linguistic analysis of the learner's response to exercises
- a Learner model that creates a dynamic model of the learner, representing his learning history
- a Tutor model that controls the system's reactions to the learner's input and the structure of materials offered.
The Way Ahead
It is too early to appreciate the real contributions of RECALL. A great deal of serious work has been done, but the results so far make identification of a promising product difficult. The consortium will focus on a clear organisation and analysis of the tasks needed to implement intelligent error recognition and diagnosis and a self-assessment of the promise of the direction taken in this project, including presentation of the various obstacles encountered. These might take the form of commentary on the current system design. The design goals of each component and the pros and cons of the current choices, and the resources needed to provide complete error correction coverage for one useful level.
The consortium is optimistic about the significance of the theoretical work done on error analysis, but the ultimate value of the project lies in the degree to which it contributes to the development of systems capable of intelligent error recognition and feedback. The partners will therefore focus on demonstrating the path to intelligent error-handling systems, and contributing to understanding of the technical hurdles to be cleared.
The demand for software for Computer-Assisted Language Learning (CALL) is increasing considerably. This demand is driven by a growing need for foreign language skills in Europe and the particular feature of CALL in providing classroom and schedule-independent teaching devices. A drawback of most currently available CALL software, however, is that it cannot fully recognise the learner's mistakes and provide sophisticated feedback with respect to appropriate remedial or revision materials.

Progress and Results

The project encompasses preparatory activities and small-scale demonstration and verification. It assumes, as its platform and starting point, a commercial CALL system currently under development at the IBM Scientific Centre in Heidelberg, the CoALA (Computer Aided Language Acquisition) system. Within this preparatory stage, the project will investigate user requirements and market opportunities for an enhanced CALL product, with the focus on detection and correction of spelling, morphological, and syntactic errors. Semantic information, world knowledge, and inference only play a role to the extent that they are necessary to detect and diagnose these errors. Toward the end of this project, a small-scale demonstrator will sample the findings in the teaching of English and German and will be tested by several user organisations involved in or connected with the consortium.

Exploitation

Due to the preparatory nature of this project, the demonstrator will not cover the full functionality of the various components, but will be a showpiece to raise awareness concerning the inherent potential of "intelligent" CALL. It is intended to use this and other results of the project to specify a follow-up project, undertaking the full implementation and validation of the system. The follow-up project would also extend into distance learning.

There are immediate business interests for the partners to engage in this area. The increasingly positive market opportunities for CALL products are an obvious motivation for IBM as a software manufacturer. For Exlingua International as a provider of linguistic education the aim is to provide language tuition, adjusted to fit the needs of the market and to do so at competitive rates and with the help of the most recent technology. Members of the user group have closely related interests either economically motivated (as publishers of electronic media) or as providers of cultural and educational services.

Funding Scheme

CSC - Cost-sharing contracts

Coordinator

IBM Wissenschaftliches Zentrum
Address
Vangerowstr. 18
69115 Heidelberg
Germany

Participants (3)

Exlingua International
United Kingdom
Address
92 Lisburn Road
BT9 6AG Belfast
UNIVERSITAET OSNABRUECK
Germany
Address
100,Pascalstrasse
49069 Osnabrueck
UNIVERSITY OF ULSTER
United Kingdom
Address
Cromore Road
BT52 1SA Coleraine