Alternative non-BERT model choices for the textual classification in low-resource languages and environments