Institutional Repository of Vilnius University
Anglų kalbos valstybinių brandos egzaminų tekstų ir užduočių klasifikavimas pagal CEFR lygius taikant mašininio mokymosi metodus /
Abstract
dc:descriptionThis study examines the classification of texts and questions from the national English proficiency exams according to CEFR levels using machine learning methods. The aim of this study is to develop and compare machine learning models designed for the automated determination of CEFR levels for texts and questions. A dataset of exams from 20202024 was collected for the study, comprising texts, questions, answers, and additional statistical features, which was later expanded. TFIDF and embedding methods were applied to convert textual data into numerical features, using the ”SentenceBERT,” ”BERTbaseuncased,” and ”RoBERTabase” language models. To solve the classification task, Naive Bayes, logistic regression, linear SVM, random forests, Knearest neighbors, and multilayer perceptron models were applied. Model performance was evaluated using accuracy, precision, recall, and F1 scores. The results of the study showed that models based on embedding features achieved better results than those based on TFIDF, and the inclusion of additional statistical features improved the quality of question classification. The results of this study indicate that machine learning methods can be used to automatically assess the level of task difficulty in English language exams and to determine CEFR levels.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ambraziejūtė, Simona,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308333910&prefLang=en_US