🧠 Open Multi-Label ASJC Classification

We present the first multi-label classification model built on the ASJC taxonomy that reliably assigns subject categories to individual documents—including those published in general-science or interdisciplinary journals—using Title, Container Title, and Abstract metadata.

👥 Team

• Michael Gusenbauer – Johannes Kepler University Linz | ORCID: https://orcid.org/0000-0001-7768-2351
• Jochen Endermann – University of Applied Sciences Kufstein
• Harald Huber – University of Applied Sciences Kufstein
• Simon Strasser – University of Applied Sciences Kufstein
• Andreas-Nizar Granitzer – Norwegian Geotechnical Institute | ORCID: https://orcid.org/0000-0002-5839-4300
• Thomas Ströhle – Universität Innsbruck | ORCID: https://orcid.org/0000-0002-1954-6412

🎯 Purpose

Traditional ASJC classification approaches are limited by incomplete sources, journal-level labels, or single-label assignments. This project provides:

• Multi-label classification across 307 subjects (compare google sheet for all labels)
• Fine-tuned SciBERT model trained on Crossref metadata
• Methods for collection-level analysis (researcher portfolios, institutions, datasets)

✨ Features

• High performance
• Works with or without source title metadata
• Open, reproducible, and ready for research use

🗂 Content

• Fine-tuned model
• Sample code for model inference

📖 Citation

If you use this work, please cite:

@article{Gusenbauer.2025,
author = {Gusenbauer, Michael and Endermann, Jochen and Huber, Harald and Strasser, Simon and Granitzer, Andreas-Nizar and Ströhle, Thomas},
year = {2025},
title = {Fine-tuning SciBERT to enable ASJC-based assessments of the disciplinary orientation of research collections},
keywords = {All Science Journal Classification;Disciplinary coverage;Fine-tuning;multi-label classification;SciBERT;Transformer-based language models},
issn = {0138-9130},
journal = {Scientometrics},
doi = {10.1007/s11192-025-05490-0},
}