Beyond reading jumbled words: Bridging perception and language in the Noisy Channel model

点击翻译按钮获取中文摘要

Classic computational models of visual word recognition do not consider the noise present in early perceptual processes, and they cannot cope with “jubmled wrods”- words with distorted letter order, unlike skilled readers. Previous work has developed the Noisy Channel model which can recognise such words, modelled as an optimal Bayesian inference process operating on a noisy visual input where there is uncertainty in the identity and order of letters. In this project, using computational modeling and behavioural experiments, the scope of the Noisy Channel model will be extended to address the role of phonology in the early stages of reading. The outcome will be a better understanding of the link between visual perception and language.