Research Paper · arXiv Preprint
SCALE: Scientific Concept Aggregation via LLMs and Embeddings
Extending the OpenAlex taxonomy with 114,000 fine-grained scientific Concepts.
Co-authored with Feichi Lu, Oliver Grun, Mariia Eremina and Andrea Perlato at MDPI Data Intelligence. SCALE extends the OpenAlex taxonomy with a new fifth level of 113,892 fine-grained scientific Concepts, built by combining SPECTER2 embeddings, LLM-based granularity classification, and Leiden graph clustering over nearly three million MDPI author keywords. The resulting taxonomy powers Atlantis and the Taxonomy Explorer, and is already deployed in production to tag MDPI's corpus of close to two million papers.
