A BERT-based Distractor Generation Scheme with Multi-tasking and Negative Answer Training Strategies
This work addresses the need for high-quality, multiple distractor generation in MCQ preparation, representing an incremental improvement over existing methods.
The paper tackled the problem of improving distractor generation for multiple-choice questions by addressing limitations in quality and single-distractor focus, resulting in a model that advanced the state-of-the-art BLEU 1 score from 28.65 to 39.81 and generated diverse, effective distractors.
In this paper, we investigate the following two limitations for the existing distractor generation (DG) methods. First, the quality of the existing DG methods are still far from practical use. There is still room for DG quality improvement. Second, the existing DG designs are mainly for single distractor generation. However, for practical MCQ preparation, multiple distractors are desired. Aiming at these goals, in this paper, we present a new distractor generation scheme with multi-tasking and negative answer training strategies for effectively generating \textit{multiple} distractors. The experimental results show that (1) our model advances the state-of-the-art result from 28.65 to 39.81 (BLEU 1 score) and (2) the generated multiple distractors are diverse and show strong distracting power for multiple choice question.