Evidence about difficult boundaries belongs to this taxonomy version.
Confused pairs, recurring out-of-scope inputs, weak definitions, and likely overlap are evaluated without changing the source concepts.
OpenAlex research topics
Classifies what a scholarly work is about. Use it for titles, abstracts, grant descriptions, and local research-area labels.
- Version
- 15 July 2026 snapshot
- Concepts
- 4,516
- Source
- OpenAlex
- Unit being classified
- The primary research topic of a scholarly work.
- Canonical output
- One stable concept ID and its published external label.
- Useful input
- Research title, abstract, grant description, or local research-area label.
- Important limitation
- A multidisciplinary work may support several topics even when a downstream workflow requires one primary concept.
What this taxonomy distinguishes
OpenAlex topics are organized under subfields, fields, and broad domains. A short title may retrieve several related topics; an abstract often supplies the method, population, or application needed to distinguish them.
Example mapping
Source value
"artificial intelligence for healthcare diagnosis"
Canonical conceptT11396 Artificial Intelligence in Healthcare
Concepts
A sample of the stable IDs, external labels, and parent context in this version.
| Concept ID | External label | Parent context | Status |
|---|---|---|---|
T11396 | Artificial Intelligence in Healthcare | Health Information Management | Active |
T10001 | Geological and Geochemical Analysis | Geophysics | Active |
T10004 | Soil Carbon and Nitrogen Dynamics | Soil Science | Active |
T10007 | Monetary Policy and Economic Impact | Economics and Finance | Active |
Grouping proposals remain separate from the published source.
Canonmap can propose reporting groups and lower-level distinctions for a named goal. Adoption is explicit; no recommendation silently changes normalization.
Production decisions identify an immutable mapping build.
The public demonstration serves an explicitly selected calibrated build. Normalize and Performance identify the exact mapping build and evaluation evidence used for requests.