A BERT-based Distractor Generation Scheme with Multi-tasking and Negative Answer Training Strategies

TL;DR

This paper introduces a BERT-based distractor generation method employing multi-tasking and negative answer training strategies to produce multiple, diverse distractors with improved quality for multiple-choice questions.

Contribution

It proposes a novel distractor generation scheme that enhances quality and supports multiple distractors, addressing limitations of prior single-distractor methods.

Findings

BLEU 1 score improved from 28.65 to 39.81

Generated distractors are diverse and effective

Model advances state-of-the-art performance

Abstract

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.