Оцінка ефективності нейромережевої системи для категоризації текстових документів

Abstract

This paper presents an evaluation of the DistilBERT model’s effectiveness for categorizing Ukrainian text documents. DistilBERT, a streamlined version of BERT, aims to retain the original's performance with reduced size and increased speed. This study focuses on the model's application for classifying texts into legal and non-legal categories using publicly available data, including court decisions and social media posts. The training encompassed several epochs, enhancing the model's adaptation to data peculiarities. The results, including high accuracy and precision metrics, affirm DistilBERT’s efficacy in this context. This research highlights the potential of neural network systems for automating the processing and categorization of Ukrainian texts in various fields.