View article

Applications that adopt peer instruction and active learning make use of short student answers as a learning opportunity: they take some answers as topics of discussion in class, or select answer rationales from peers to foster self-reflection on the learner’s own rationale to a question. However, experience with these applications reveals that some answers are irrelevant, inappropriate, or could even be offensive, and must therefore be filtered out. Automatic identification of these rationales is the topic of this research. We introduce an easy-to-implement approach based on standard text classification techniques, namely bag of words and vector space models, and show its effectiveness for filtering irrelevant answers.