In this paper, we present the SEJO KOST 0.1, a new language resource that extends the KOST corpus of written texts produced by learners of Slovene as a non-first language. Compared to the KOST corpus, the new resource introduces two key innovations: more balanced sampling with respect to the learners’ first language and the assignment of a CEFR level to each text. SEJO KOST is still under development. We describe the design of the subcorpus, the collection of texts, the procedure for assigning CEFR levels, and the current Version 0.1. In conclusion, we outline directions for future work.
|