The Design of Tests with Multiple Choice Questions Automatically Generated from Essays in a Learner Corpus

Learner corpora have great potential as sources of educational material. If a corpus contains annotations of mistakes in student works, it can be of use for the recognition and analysis of the most common error patterns. The error-annotation system of the learner corpus REALEC makes it possible to automatically generate different types of test questions and thus form exercises from the corpus data. This paper describes the creation of an automatic multiple-choice generator which works with the specific types of the student errors annotated in the texts of examination essays