Home People Projects Talks and Publications Events

Pre-Workshop courses on

Interfaces of Intelligent Computer-Assisted Language Learning
The Ohio State University, December 14-16, 2006


(The program for the ICALL 2006 workshop on December 17 can be found here.)


Unless otherwise noted, all events take place in Room 255 of Hagerty Hall on the main OSU campus (1775 College Rd. Columbus, Ohio 43210).

December 14 - Pre-Workshop Course 1

Kathleen McCoy: Learner Modeling and NLP in ICALL

9:30 - 10:00  Coffee

10:00 - 12:00  Session 1

12:00 - 1:00  Lunch

1:00 - 3:00  Session 2

3:00 - 3:30  Coffee break

3:30 - 5:30  Session 3

December 15 - Pre-Workshop Course 2

Eckhard Bick: VISL - An Integrated Multi-lingual Approach to CALL

9:00 - 9:30  Coffee

9:30 - 11:30  Session 1

11:30 - 1:00  Lunch

1:00 - 3:00  Session 2

3:00 - 3:30  Coffee break

3:30 - 5:30  Lab Session (Note: This session takes place in Derby Hall, Room 29.)

December 16 - Pre-Workshop Course 3

Susan Bull: Open Learner Models

9:00 - 9:30  Coffee

9:30 - 12:00  Session 1 (second part of session in lab: Hagerty Hall, Room 171)

12:00 - 1:30  Lunch

1:30 - 4:00  Session 2 (including Coffee Break tbd.)

4:00 - ...  Pizza in Hagerty Hall, Room 255. Time for individual meetings and discussions.


NLP and Learner Modeling for ICALL

Kathleen McCoy (University of Delaware)

An important problem in ICALL is modeling the knowledge that the learner has of the language being learned. Learner modeling is important for both effectively diagnosing the underlying causes of a studentís displayed errors, and for effectively tutoring the student once an error has been identified. Issues in learner modeling span the spectrum of (1) what is to be modeled (e.g., the learnerís grammar of the language being learned, the learnerís objective knowledge of the language, the learnerís knowledge of other languages that may affect their production in the language being learned, information about the learnerís approach to the task at hand such as whether they are careless), (2) how the model is initialized, (3) how and when the model is updated over time, (4) what effects the learner model has on the system.

This course will survey the literature on learner modeling in ICALL and explore both current practices and open questions. Emphasis will be placed on how natural language processing techniques can be leveraged in learner modeling. The learner model in the ICICLE system (a model of the learnerís grammar of the language being learned) will be used to motivate some of the open issues.

VISL - An Integrated Multi-lingual Approach to ICALL

Eckhard Bick (Southern Denmark University)

This course will present tools and ideas developed by the VISL project at the University of Southern Denmark. The original focus of VISL's work in the CALL arena was grammar and syntax learning in a cross-language perspective, where a unified descriptive system of grammar was implemented using interactive web-based tools like grammar games and a syntactic tree-building tool, supported by small, linguist-annotated sentence collections (treebanks) for 26 languages.

In its current research efforts, VISL is designing NLP software (taggers and parsers) for a number of these languages in order to add a layer of linguistic "live" intelligence to the system. Many aspects central to active language learning, like variability and naturalness, as well as user-driven rule deduction and exploration, will only work in a CALL environment if it has access to NLP tools and annotated corpora (such as our 8-language collection at http://corp.hum.sdu.dk). Though this particular aspect of the "I" in "ICALL" (NLP) is of course much harder to address than the "C", doing so will facilitate a broader interpretation of grammar teaching as language awareness building, and it will also allow the integration of secondary learning aids, like grammar-marked spellchecking or even machine translation.

If time permits and if participants express their interest in this, a practical demonstration and/or corpus exercises will be part of the course, drawing on VISL's annotated corpora and didactic concepts developed for the teaching of language awareness in Danish schools and universities.

Open Learner Models

Susan Bull (University of Birmingham)

In most adaptive learning environments the learner model contents are not directly presented to the user. Open learner models are learner models that are in some way accessible - usually to the learner being modelled. This may be for viewing only, or it may be interactive, with the learner having varying levels of control over the content of their learner model. Some environments also allow learner models to be opened to others, such as instructors, parents or peers. Reasons for opening the learner model are varied, but include promoting learner reflection, facilitating navigation to information, supporting collaboration, and allowing the learner to contribute data to their learner model to improve system adaptation.

This course will introduce current research in open learner modelling, including simple and complex examples of learner models that can be presented to the user, and that allow interaction about the learner model data. After a general introduction to the field, issues specific to open learner modelling in language learning will be considered. A lab session will allow participants to experience some currently deployed open learner models.

Last modified: December 11, 2006