Dyslexia is a type of learning impairment that affects 5–15% of the world’s population. MusVis, a web game developed by Maria Rauschenberger and supervised by Ricardo Baeza-Yates and Luz Rello, researchers affiliated with UPF’s Department of Information and Communication Technologies (DTIC), won the W4A Attendees’ Award on 20 April at the 17th International Web for All Conference, whose specific theme this year was Automation for Accessibility, for the communication paper entitled “Screening risk of dyslexia through a web-game using language-independent content and machine learning”.
“To our knowledge, this is the first time the risk of dyslexia has been analysed by means of a web game based on language-independent contents and using machine learning”, Rauschenberger affirms. This new method could be used to detect possible learning disorders in children, even before they develop language skills, and lead to possible early intervention. Thus, “we aimed to detect dyslexia through interactions that do not require a knowledge of language”, the authors affirm.
The team created the game material by analyzing the mistakes made by persons with dyslexia in several languages, as well as other dyslexia-related characteristics such as auditory and visual perception. They conducted a user research with 313 youngsters (116 of whom were dyslexic) and used the obtained data to build predictive machine learning models. Using Random Forests and Extra Trees, the approach achieves an accuracy of 0.74 for German and 0.69 for Spanish, as well as an F1-score of 0.75 for German and 0.75 for Spanish.
Although the discrepancies are unlikely to be as significant or obvious as the reading and spelling mistakes that define children with dyslexia, the authors believe MusVis is a potential technique for detecting dyslexia in pre-readers using language-independent audio and visual material. “Because children with dyslexia require around two years to overcome their challenges, our technique, because it is language-independent, might help decrease school failure, delayed treatment, and, most significantly, alleviate the suffering of children and parents,” Rauschenberg emphasizes.
“Our approach might optimize the resources to detect and treat dyslexia, however, we would need to examine many more children at an early age to expand the training data for our predictive models based on machine learning and improve our results”, the authors add.
Screening risk of dyslexia through a web-game using language-independent content and machine learning, Maria Rauschenberger, Ricardo Baeza-Yates, Luz Rello
Published: April 2020