Exploring transcription processes when children with and without reading and writing difficulties produce written text using speech recognition

Abstract

The aim of this study was to investigate composition and error-correction processes, and their relationship with production rate, in children, age 10-12, with and without reading and writing difficulties using speech-to-text (STT) to write expository texts in Swedish. Measures of individual abilities: working memory, spelling ability and decoding ability, and the ability to interact with the STT tool under optimal conditions (STT success rate) were collected. For both those with and without difficulties, neither working memory, nor spelling or decoding ability predicted burst length nor accuracy. Only a child’s STT success rate did predict accuracy during text composition. Further, none of the individual abilities predicted choice of error-correction modality (keyboard or STT) or error correction functionality. This indicates that the children’s behavior were independent of these abilities. Furthermore, production rate was significantly predicted by both burst length and accuracy, and by working memory, but not by error-correction behaviour, nor by spelling or decoding ability. This indicates that composing text using STT is a cognitively complex process placing heavy demands on working memory. Dictating more than one word at a time and combining STT and keyboard use were identified as two useful strategies that can be taught in STT instruction.