Sahinde Dogruer, Abdallah AlShafey, Mohamed Mohamed, Alexander Timofeev
We developed a model that evaluates the language proficiency of 8th-12th grade English Language Learners (ELLs) based on 6 criteria, namely grammar, vocabulary, syntax, cohesion, conventions, and phraseology. The evaluation metric of the model is chosen to be the MCRMSE, or the mean column-wise root mean squared error, where each column corresponds to one of the 6 criteria being evaluated. Our model scored a MCRMSE of ≈ 0.5, performing significantly better than our human-baseline model. We also provide an easy-to-use interactive demo targeting the potential end users.