Bayesian Modelling of Confusability of Phoneme-Grapheme Connections

Deficiencies in the ability to map letters to sounds are currently considered to be the most likely early signs of dyslexia [4]. This has motivated the use of Literate, a computer game for training this skill, in several Finnish schools and households as a tool in the early prevention of reading disability. In this paper, we present a Bayesian model that uses a student’s performance in a game like Literate to infer which phoneme-grapheme connections student currently confuses with each other. This information can be used to adapt the game to a particular student’s skills as well as to provide information about the student’s learning progress to their parents and teachers. We apply our model to empirical data collected using Literate. Based on these data, we evaluate and compare using Bayesian methodology three different submodels with different restrictions on the possible confusability relations.

• Type: Conference
• Year: 2007
• Where: ICALT
• Authors: Mikko Vilenius, Janne V. Kujala, Ulla Richardson, Heikki Lyytinen, Toshio Okamoto
• Keywords: ICALT 2007, Likely Early Signs, Machine Learning, Phoneme-grapheme Connections Student, Student’s Learning Progress