Semi-automatic Test Generation for Tandem Learning

We introduce a Web-based CALL architecture that facilitates the construction of learner-customized multiple choice tests in a cross-lingual tandem language learning environment. Mistakes made by the learner are manually corrected and classified by his tandem partner, who acts as a tutor. If the learner has problems to identify and correct his mistakes, or if he likes to practice, he can generate two kinds of tests that are automatically produced on the basis of the tutor’s error classifications. We describe the NLP components behind our test generator, give a preliminary evaluation of the quality of the generated tests, and present the current state of our Web application.