December 4, 2025ScientometricsOpen Access

Fine-tuning SciBERT to enable ASJC-based assessments of the disciplinary orientation of research collections

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Analysis reveals improved multi-label classification accuracy in research collections, indicating effective ASJC taxonomy use.

Achieved a weighted F1-score of 0.892 on 307 subjects in multi-label classification.
Fine-tuned SciBERT model trained on a Crossref dataset for enhanced subject classification accuracy.
Introduced label-averaging method for assessing disciplinary orientation and document comparison.
Model published on Hugging Face for greater accessibility and reproducibility in ASJC classification.

Harald Huber (2025) studied this question.

View Full Paper

Ask AI

Bookmark