Year of Graduation
Evaluating and Improvement of the Annotation of RLC
Fundamental and Computational Linguistics
This work is devoted to inter-annotator agreement between annotators of Russian Learner Corpus. The problem of annotator coherence is common for different corpora, especially new ones. By examining that important feature, I ensure that all errors are marked correctly. For RLC this examination is crucial because it is held for the first time. For measuring consistency, I use Cohen’s kappa coefficient. Data have been collected from several texts from the corpus. Each text is annotated by two people independently, so in the end there are two variants for each of 50 files. After that, the program, which was created using Python program language, is used for calculating Cohens’s kappa coefficient. Results of this calculation show what and how should be improved in annotators’ instructions.