Predicting Learner Levels for Online Exercises of Hebrew

We develop a system for predicting the level of language learners, using only a small amount of targeted language data. In particular, we focus on learners of Hebrew and predict level based on restricted placement exam exercises. As with many language teaching situations, a major problem is data sparsity, which we account for in our feature selection, learning algorithm, and in the setup. Specifically, we define a two-phase classification process, isolating individual errors and linguistic constructions which are then aggregated into a second phase; such a two-step process allows for easy integration of other exercises and features in the future. The aggregation of information also allows us to smooth over sparse features.