Reading skill is essential for academic success, but many children fall behind and never catch up in reading skill due to learning differences (such as dyslexia) or opportunity differences (poverty). Further, changes in educational practices and policies and innovations in technology have not had a demonstrable effect on reading skill over the past 20 years and have not closed gaps in reading skill for children with learning differences or children from lower socioeconomic strata or children of color.
The project proposes to create an open database with unique educational value for improving reading outcomes in US children. The database will include annotated speech samples taken from kindergartners being screened for risk for future reading difficulty and the ground truth of whether those children become good or poor readers in 3d grade. This will allow Kaggle competitions to (1) have speech recognition make in-school screening of all children be efficient, valid, scalable, and equitable, and (2) predict from kindergarten speech data which children progress to good or poor reading (so that immediate support can be given to children otherwise fated to struggle with reading).