28  Last view: 2025-05-222  Last update: 2024-10-24

EKI Estonian L2 learner parallel error corpus

The corpus has been automatically generated from the Error-annotated Estonian L2 learner corpus. The materials for the error annotated corpus are based on the Estonian learner corpus EMMA, containing an Estonian learner assessment test (7th grade, 254 texts), basic school final exam (9th grade, 251 texts) and state exam data (12th grade, 998 texts) from the Education and Youth Board. The error-annotation has been carried out by applying an ERRANT-M2 annotation scheme, adapted to the Estonian specifics. The corpus includes for all of the original and two reference sets a total of 1,503 essays of A2, B1 and B2 level in subsets of 1202-150-151 for train, dev and test. The reference sets are differentiated by the basis of whether optional error corrections are taken into account or not.