Generating Modular Grammar Exercises with Finite-State Transducers

This paper presents an ICALL system for learning complex inflection systems, based upon finite state transducers (FST). Using a FST has several advantages: it makes it possible to generate a virtually unlimited set of exercises with a relatively small amount of work, and it makes it possible to process both input and output according to a wide range of parameters, such as dialect variation, and varying writing conventions. It also makes it possible to anticipate common error types, and give precise feedback both on errors and possible corrections. It shifts the developer’s focus from form generation and over to a pedagogically-motivated modelling of the learning task. The system is in active use on the web for two Saami languages, but can be made to work for any inflectional language.