Predicting CEFR levels in learners of English: The use of microsystem criterial features in a machine learning approach

Abstract This paper focuses on automatically assessing language proficiency levels according to linguistic complexity in learner English. We implement a supervised learning approach as part of an automatic essay scoring system. The objective is uncover Common European Framework Reference for Languages (CEFR) criterial features writings by learners English foreign language. Our method relies the concept microsystems with related learner-specific systems which several forms operate paradigmatically. Results internal data show that different help classify from A1 C2 (82% balanced accuracy). Overall results external combination lexical, syntactic, cohesive and accuracy yields most efficient classification across corpora (59.2%