A Novel Machine Learning Approach to Generating Mind Maps For Visual Learners

Abstract

Many learners with developmental disabilities, such as dyslexia, aphasia, and autism tend to be visual learners, who learn better when given visual aids in conjunction with written text. A similar benefit is seen for English Language Learners as well. Unfortunately, there is a key lack of methods available to transform vital academic text into a visual format for such learners, with current methods primarily focusing on text-to-text generation, e.g. summarization. My approach, named MindTree, fixes this by automatically generating informative mind maps for any length of textbook or article text. MindTree picks out the key topics from long and complicated texts and organizes them in a hierarchical and logical mind map, drawing connections between related topics. My approach additionally finds latent, or “hidden” topics within the text that may not be explicitly mentioned. In this paper, I investigate the effectiveness of MindTree-generated mind maps using objective and subjective measurements for improving the accessibility and learning outcomes of academic text for learners with reading disabilities.