General

ICALL - Intelligent Computer Assisted Language Learning

“A System Architecture for Intelligent Computer-Assisted Language Learning“ is a project funded by NordPlus Sprog for two years (2011-2013). It is run by three partners:

• Reykjavik University , Hrafn Loftsson
• University of Gothenburg, Lars Borin
• University of Iceland, Birna Arnbjörnsdóttir

The aim of the project is to develop an open-source system architecture for supporting ICALL, i.e. CALL that reuses NLP tools and NL resources, with emphasis on the Nordic languages. The main goal is to make the architecture as language independent as possible such that underlying relevant NLP tools can be easily plugged in as external modules depending on the language in question. The same holds for the type of activities (exercises) that the architecture will be able to support, i.e. new activities will be “plug and play” to as large extent as possible.

Specific for Swedish

In designing the new architecture for the Swedish application we intend first to port the existing Swedish FoF (Focus-on-Form) exercises developed for the ITG application and add the Swedish vocabulary exercises developed previously at our department. Having the existing ITG exercises will allow us to quickly assess the viability of the architecture for this kind of applications. Together with the new modules to be developed in this project, they make up a broad and varied spectrum of ICALL applications which will allow us to test the flexibility of the architecture. The ITG exercises use manually annotated corpora and although the text material is authentic, it is also now slightly dated and becoming more so all the time. One goal of this project is thus to adapt the language tools at our disposal with the aim of achieving the same kind of functionality using arbitrary text, e.g. from the internet. Another goal is to extend the range of FoF exercises offered and to explore how these exercises should be connected to other language learning activity types.

The upcoming ICALL platform for Swedish is under active development and has the working title “Lärka” - “LÄR språket via KorpusAnalys”.

The work on Lärka comprises three central parts:

1. conversion of two existing exercise generators for training Swedish grammar and vocabulary
2. from two different idiosyncratic, difficult-to-maintain, monolithic programs into a modular web-based framework already used for our general corpus infrastructure, and
3. at the same time developing and refining both the exercise generating functionality and the user interaction aspect.

While (1) and (2) above are straightforward programming tasks, (3) requires more attention:

The existing applications use manually annotated Swedish standard corpora as the source of the authentic text passages out of which the exercises are constructed: Stockholm Umeå Corpus (1 MW; parts of speech and vocabulary) and Talbanken (75 kW; syntactic analysis). We need to ensure that only appropriate passages are selected for exercises. This, in our view, is reduced to two main points: readability analysis and ranking of the examples. The readability analysis of the source texts should ensure that training contexts for test items are appropriately chosen for the target learner proficiency level. Readability is a complex notion, however, and work on finding an appropriate measure is ongoing. Ranking the passages is an interesting problem, since the notion of ”good examples” is often subjective and sometimes conflicts with the notion of ”authentic examples”. User tracking, including avoiding repetition of examples, is another challenge for a system without a login procedure.

How should we view our application - ”intelligent tutor” or support to teachers, i.e. should it be totally automatic or allow some user editing of exercise items? On the one hand, it is natural to expect teachers to be critical to automatically generated items. On the other hand, we expect mainly students to use our service. So the question is - who will be the main user? Can we meet the needs of both user groups? Depending on our answers to these questions the application will need different functionalities.

On the user interface side we have to decide how to present exercises – by learner proficiency level, by exercise type, etc. This is not only a question of user-friendliness, but also of potential to extend the system and expand its scope.

The first version of the application is planned for August 2012 with five types of exercises: grammar exercises (part-of-speech and syntactic analysis); and vocabulary exercises (wordbank items, multiple choice items and c-tests ).

Publications

• Elena Volodina, Lars Borin, Hrafn Loftsson, Birna Arnbjörnsdóttir, Guðmundur Örn Leifsson. 2012. Waste not, want not: Towards a system architecture for ICALL based on NLP component re-use. Workshop on NLP in Computer-Assisted Language Learning. [pdf]
• Elena Volodina, Hrafn Loftsson, Birna Arnbjörnsdóttir, Lars Borin, and Guðmundur Örn Leifsson. 2012. Towards a system architecture for ICALL. In G. Biswas et al. (eds), Proceedings of the 20th International Conference on Computers in Education. Singapore: Asia-Pacific Society for Computers in Education. [pdf]
• Elena Volodina and Lars Borin. 2012. Developing a freely available web-based exercise generator for Swedish. EuroCALL 2012 Proceedings, Gothenburg. [pdf]