Research category classification of scientific articles on human health risks of electromagnetic fields using pre-trained BERT

This paper presents bidirectional encoder representations from transformers (BERT)-based deep learning model for the classification of scientific articles. This model aims to increase the efficiency and reliability of human health risk assessments related to electromagnetic fields (EMF). The proposed model takes the title and abstract of EMF-related articles and classifies them into four categories: animal exposure experiment, cell exposure experiment, human exposure experiment, and epidemiological study. We conducted a performance evaluation to verify the superiority of the proposed model. The results demonstrated that the proposed model outperforms other deep learning models that use pre-trained embeddings, with an average accuracy of 98.33%.