Low-resource text classification using cross-lingual models for bullying detection in the ukrainian language

Abstract

The object of research is multilingual models for training on limited datasets. The article reviews multilingual models for training on limited datasets and analyzes their development. Multilingual models are used for many low-resource languages, but Ukrainian is not one of them. The purpose of the work is to increase the effectiveness of text classification in the conditions of a limited set of data in the Ukrainian language by using multilingual models, a zero-shot learning approach i.e. without a target language, and using machine translation to create or augment a dataset. Ref. 24, pic. 5, tabl. 3

Keywords

• multilingual models
• zero-shot classification
• bullying detection
• xlmroberta
• mbert
• laser
• muse