Automatic error annotation in learner corpora (on the material of Ukrainian learner English corpus)

Having analysed the most common learner mistakes we made an attempt to establish more rigid boundaries between the error categories and subcategories. For each error detection level - grammar-based vs formal errors - we have developed a broader taxonomy, having established the frequency of each error type and its influence on the performance of automatic annotation tools. Being compared with the corresponding results for other learner corpora as well as native language corpora, these data can become the basis for extracting the distinctive features of English texts produced by Ukrainian speakers. Therefore, we view the potential for further research in the contrastive analysis of the texts produced by language learners from different linguistic and ethnic backgrounds and developing NLP tools for native.
The object of the research: the performance of automatic corpus annotation tools on Ukrainian learner English corpus.
The subject of the research: the peculiarities of Ukrainian learner English and their influence on the accuracy rate of the performance of automatic corpus annotation tools.

Hupalyk I. A. Automatic error annotation in learner corpora (on the material of Ukrainian learner English corpus) : Master Thesis : 035 Philology / Hupalyk Iryna Andriivna. - Kyiv, 2021. - 71 с.