The Classification of Short Scientific Texts Using Pretrained BERT Model.

Area of Science:

• Biomedical Informatics
• Natural Language Processing

Background:

• Scientific literature selection is crucial but time-consuming.
• Automated text classification offers a potential solution.

Purpose of the Study:

• To evaluate advanced natural language processing models for classifying scientific abstracts.
• To compare the performance of PubMedBERT and ensemble models on a specific biomedical dataset.

Main Methods:

• A dataset of 630 PubMed abstracts was used for binary classification.
• Twenty-seven PubMedBERT model variations and four ensemble models were tested.
• Three hundred resampled tests were conducted for each approach.

Main Results:

• The best PubMedBERT model achieved an F1-score of 0.857.
• The best ensemble model reached an F1-score of 0.853.
• Both approaches demonstrated high efficacy in text classification.

Conclusions:

• State-of-the-art natural language processing models can enhance the quality of scientific text classification.
• Automated classification significantly aids in efficient scientific literature selection.