TIB/ORKG-core-domain-classifier-dataset

ORKG Core Domain Classifier — Metadata

Update Date: 2026-10-21

Research-domain labels produced by the ORKG Core Domain Classifier — one row per document.

The classifier that produces this file lives at https://gitlab.com/TIBHannover/orkg/nlp/experiments/core-domain-classifier. This card is generated, so please do not edit it by hand.

Load the dataset

from datasets import load_dataset

dataset = load_dataset("TIB/ORKG-core-domain-classifier-dataset", split="full")

print(dataset[0])
# {'id': '100000028', 'predicted_domain': 'Biomedical Engineering and Bioengineering'}

Schema

Column	Description
id	document id
predicted_domain	research domain assigned by the classifier

Get the title, abstract and details of a paper

The dataset only stores ids and labels. The ids are CORE work ids, so the full record can be fetched from the public ORKG Ask index (no key needed):

import requests

paper_id = "103862900"
r = requests.get(f"https://api.ask.orkg.org/index/get/{paper_id}",
                 params={"fields": "id,title,abstract,year,authors,doi"})
print(r.json()["payload"])

Example response (abstract shortened). The response also contains other fields such as publisher, journal and urls. They are null when the index has no value for them:

{
  "id": "103862900",
  "title": "Towards a unifying visualization ontology",
  "abstract": "Although many terminologies, taxonomies and also first ontologies for visualization have been suggested, there is still no unified and formal knowledge representation ...",
  "year": 2011,
  "doi": null,
  "authors": [
    "M. Voigt",
    "J. Polowinski",
    "Technische Berichte",
    "Fakultät Informatik",
    "Technische Universität Dresden"
  ]
}

The source repository also ships a helper (app/paper_lookup.py) that looks up many ids in parallel.

How the labels were produced

Each label is the output of an LLM prompted with the document's title, abstract and topics. They are model predictions, not human annotations, and there is no gold-standard evaluation set. Labels are drawn from a fixed vocabulary of research domains, published as app/assets/domains.json in the source repository.

📊 Current Stats — v4 (2026-10-21)

Total Classified Records: 47,716,392

Classification Label	Count	Percentage (%)
Arts and Humanities	4,062,072	8.51%
Engineering	1,495,261	3.13%
Life Sciences	1,311,100	2.75%
Physical Sciences and Mathematics	666,404	1.40%
Social and Behavioral Sciences	10,440,745	21.88%
Aerospace Engineering	404,571	0.85%
Biomedical Engineering and Bioengineering	4,444,307	9.31%
Chemical Engineering	612,976	1.28%
Civil and Environmental Engineering	995,222	2.09%
Computer Engineering	1,375,741	2.88%
Computational Engineering	542,989	1.14%
Electrical Engineering and Information Technology	1,050,260	2.20%
Electrical and Computer Engineering	601,736	1.26%
Energy Process Engineering	266,482	0.56%
Engineering Science and Materials	797,135	1.67%
Materials Engineering	139,689	0.29%
Materials Science and Engineering	243,873	0.51%
Mechanical Engineering	209,700	0.44%
Nanoscience and Nano Technology	70,733	0.15%
Nuclear Engineering	108,652	0.23%
Operations Research, Systems Engineering and Industrial Engineering	661,838	1.39%
Production Technology	46,654	0.10%
Plastics Engineering	132,510	0.28%
American Studies	218,557	0.46%
Classics	153,548	0.32%
Comparative Literature	230,559	0.48%
Educational Research	756,221	1.58%
English Language and Literature	90,206	0.19%
Religion	324,944	0.68%
History	1,362,797	2.86%
History of Art, Architecture and Archaelogy	914,763	1.92%
Language, Societies and Cultures	146,742	0.31%
Literary Studies	427,995	0.90%
Musicology	122,223	0.26%
Philosophy	241,108	0.51%
Animal Sciences	218,888	0.46%
Agriculture, Forestry and Veterinary Medicine	771,736	1.62%
Biochemistry, Biophysics, and Structural Biology	214,918	0.45%
Bioinformatics	75,913	0.16%
Biotechnolgy	107,212	0.22%
Public Health	533,220	1.12%
Systems Biology	39,663	0.08%
Physiology	163,641	0.34%
Nutrition	48,417	0.10%
Nursing Pharmacology, Toxicology and Environmental Health	159,107	0.33%
Neuroscience and Neurobiology	377,682	0.79%
Microbiology	197,044	0.41%
Medicine	954,053	2.00%
Immunology and Infectious diseases	625,177	1.31%
Genetics and Genomics	149,700	0.31%
Forestry and Forest Sciences	207,796	0.44%
Entomology Food Science	145,395	0.30%
Ecology and Evolutionary Biology	335,846	0.70%
Cell and Developmental Biology	63,068	0.13%
Water Research	227,532	0.48%
Statistics and Probability	71,668	0.15%
Polymer Research	27,292	0.06%
Statistical Physics, Soft Matter, Biological Physics, Nonlinear Dynamics	44,316	0.09%
Physics	1,037,341	2.17%
Particles, Nuclei and Fields	58,720	0.12%
Optics, Quantum Optics and Physics of Atoms, Molecules and Plasmas	76,989	0.16%
Oceanography and Atmospheric Sciences and Meteorology	313,173	0.66%
Mineralogy, Petrology and Geochemistry	23,319	0.05%
Oceanography	23,561	0.05%
Mathematics	504,414	1.06%
Computer Sciences	829,076	1.74%
Earth Sciences	452,502	0.95%
Applied Mathematics	22,601	0.05%
Chemistry	134,248	0.28%
Chemical Solid State and Surface Research	12,095	0.03%
Astrophysics and Astronomy	284,468	0.60%
Urban Studies and Planning	257,204	0.54%
Sociology	12,487	0.03%
Social and Cultural Anthropology and Ethnology	90,799	0.19%
Social Sciences	387,144	0.81%
Political Science	277,780	0.58%
Psychology	159,445	0.33%
Public Affairs, Public Policy and Public Administration	250,227	0.52%
Linguistics	37,725	0.08%
Jurisprudence	53,308	0.11%
Economics	535,143	1.12%
Communication Sciences	314,814	0.66%
Agricultural and Resource Economics	29,142	0.06%
Language, Socieities and Cultures	109,070	0.23%

🕓 Revision History

Version	Update Date	Records	Change
v4	2026-10-21	47,716,392	+2,371,601 records vs v3
v3	2026-10-01	45,344,791	+6,061,674 records vs v2
v2	2026-09-02	39,283,117	+3,117,124 records vs v1
v1	2026-08-14	36,165,993	first release