CREG-5K r9192

Learner corpus consisting of answers to German reading comprehension questions written by American college students learning German. Along with the learner answers, the collection includesthe reading comprehension questions, and the target answers  that teachers prepare as reference for the grading process. The meaning of each learner answer is assessed by two independent annotators. Meaning assessment is done using a binary classification (correct vs. incorrect) as well as using a richer set of diagnosis categories encoding the nature of the divergence from the target answers specified by the teachers. Following Bailey and Meurers (2008), we distinguish "missing concept", "extra concept", "blend" (missing concept and extra material), and "non-answer" for answers which are unrelated to the topic under discussion.
Answers may not be rated by two annotators, but by none ore one instead, this is a full snapshot. The annotation furthermore includes an automatic question form detection which is not really reliable.