NVR Guess: Automated Question Generation for Honing NonVerbal Reasoning Skills

Abstract

Non-verbal reasoning tests allow evaluators to test a diverse set of abilities in students without relying upon, or being limited by, language skills. In this paper, we present an automated system to generate a variety of non-verbal reasoning questions spanning abstract reasoning, non-verbal analogy and spatial reasoning question types with granular control over difficulty levels. Our design also generates meaningful distractors as this is an important aspect of generating better quality multiple-choice questions. We present experimental results through analytics on question feedback and performance that demonstrate that the system-generated questions indeed serve to assess and hone nonverbal reasoning skills.