Microsoft is making a massive push into the classroom with Windows 10 S, Office 365 and Microsoft Teams for Education. Microsoft Research is also getting in on the act, developing technology which may end up in the next version Microsoft’s classroom software.
In a recent publication, Microsoft Research describes an AI-driven system which could help teachers automatically assess reading performance for students, saving them time and allowing more individual attention to students who need it the most.
Their research paper, “Automatic Evaluation of Children Reading Aloud on Sentences and Pseudowords,” automatically predicts the overall reading aloud ability of primary school children (6-10 years old), based on the reading of sentences and pseudowords.
To predict these scores automatically, features based on reading speed and number of disfluencies were extracted, after an automatic disfluency detection, based on opinions of primary school teachers previously gathered as ground truth of performance, who provided 0-5 scores closely related to the expectations at the end of each grade.
The AI-based system was almost as good as that of the teachers, and more reliable than individual assessors.
While the paper does not discuss commercialization of the technology we can easily see such a module forming part of Microsoft Teams for Classrooms.
Their research was published in Proc. Interspeech and can be read in full here.