# ALCHEM Lab knowledge graph 1.0.0

Built 2026-10-21 from alchem-kg ee3393a. Licence: CC BY 4.0 (ontology and data). Publisher: ALCHEM Lab, Hanyang University
ERICA (Byung-Hyun Kim). Contact: through https://alchemlab.hanyang.ac.kr.

## Files
- ontology.ttl — the ontology (OWL 2; owl:versionIRI https://w3id.org/alchem/ontology/1.0.0): materials,
  computations, observations, people, activities, provenance; CRediT roles and the lab's finer roles as SKOS concepts.
- data.trig — the data as named graphs, each with its provenance (graph/provenance):
  graph/bibliographic (papers, journals, authorships, people, institutions, projects, grants), graph/paper/<n> (what was
  extracted from paper n, prov:wasGeneratedBy its extraction run), graph/contributions (activities and who performed
  them, only where the paper says so), graph/identity (material kinds shared across papers; what each material mention
  refers to), graph/review (corrected or reviewed records).
- papers/<n>.ttl — one paper: its records and its bibliographic part.
- shapes.ttl — SHACL shapes the data conforms to.
- competency-questions.yaml — the questions the graph must answer, with expected, forbidden and unknown answers.
- catalog.ttl — DCAT record with checksums.

## Identifiers
- The https://w3id.org/alchem/ names are identifiers, not yet registered with w3id.org, so they do not resolve; the
  files are at https://alchemlab.hanyang.ac.kr/ontology/v1/. The names will not change when they are registered.
- https://w3id.org/alchem/paper/<n> is the paper with the site's number n; its records are paper/<n>/<id> (M1 material,
  C1 computation, V1 value, F1 finding …) and keep their ids across releases (a corrected record keeps its id).
- person/<id>, org/<OpenAlex id>, project/<slug>, grant/<number> come from the site's records; material/<k> is a material
  kind minted once in the identity registry; a merge or split is logged and the old IRI is kept as a pointer.
- Identity links: DOI, ORCID and ROR are given with rdfs:seeAlso; a shared ORCID or ROR means the same person or
  institution; an OpenAlex ID does not.

## What the data does not say
- Who did what is recorded per research group (graph/group-parts), not per person: the lab (the PI and the members the
  byline marks) or a partner group (the other authors, by the institution printed first for them). A group did a part
  when any of its authors performed, supervised or directed it; basis stated (the paper says) or inferred (two models
  judged from the whole paper). Machine-judged; no person reviewed it. An author's CRediT roles, where the paper has a
  statement, are kept as the paper prints them.
- Institutions are those printed for an author on that paper, not where the work was done.
- A material kind shared by two papers is the same kind of material, not the same sample or model.
- Extracted by AI models (two models, a third on disagreements) and checked by rules (OWL 2 RL, SHACL, competency
  queries); not yet reviewed by a person except where graph/review says so.