Analysis of typefaces designed for readers with developmental dyslexia: Insights from neural networks

Abstract

Developmental dyslexia is a specific learning disability that is characterized by severe difficulties in learning to read. Amongst various supporting technologies, there are typefaces specially designed for readers with dyslexia. Although recent research shows the effectiveness of these typefaces, the visual characteristics of these typefaces that are good for readers with dyslexia are yet to be revealed. This research aims to explore the possibilities of using neural networks to clarify the visual characteristics of Latin dyslexia typefaces and apply those to typefaces in other languages, in this case, Japanese. As a first step, we conducted simple classification tasks to see whether a CNN identifies subtle differences between Latin dyslexia typefaces and standard typefaces, and whether it can be applied to classify Japanese characters. The results show that CNNs are able to learn visual characteristics of Latin dyslexia typefaces and classify Japanese typefaces with those features. This indicates the possibility of further utilizing neural networks for research regarding typefaces for readers with dyslexia across languages.