File size: 3,043 Bytes
c23e353 a429173 d7c245d a429173 2fdc2c2 a429173 2fdc2c2 d7c245d a429173 c23e353 a429173 c23e353 a429173 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 | IRexp / IRSpectra-Bench — data licensing and attribution
=========================================================
This dataset redistributes ONLY extracted numeric data (IR band lists, 1H/13C shift
lists, and resolved structures as SMILES/SELFIES/InChIKey) together with each record's
source DOI/accession. No source full text, figures, or PDFs are redistributed.
Every record carries stamped licence fields
-------------------------------------------
After `scripts/join_pmc_licences.py` (Europe PMC join) plus Crossref empty-licence recovery (2026-08-27):
| Field | Meaning |
|---|---|
| `license` | Normalised label (CC-BY, CC-BY-NC*, CC0, CC-BY-SA-4.0, EMPTY, UNKNOWN, …) |
| `license_raw` | Europe PMC `license` string (PMC rows) or Chemotion landing-page term |
| `license_pool` | `commercial` \| `non_commercial` \| `sharealike` \| `empty_unknown` \| `other` |
| `license_source` | `europepmc` \| `chemotion` |
| `pmcid` | Normalised PMCID when source is PMC |
Full corpus: `data/irexp/irexp.jsonl.gz` (121,233 records) — **kept on disk with
per-record licence**; filter at release time.
Licence pools (record counts)
-----------------------------
| Pool | Records | Redistribution policy |
|---|---:|---|
| **commercial** (CC0 + CC-BY) | **88,545** | Primary Sci Data / Zenodo redistributable artifact |
| non_commercial (CC-BY-NC*) | 21,823 | Held aside — `licence_pools/irexp_non_commercial.jsonl.gz` |
| sharealike (Chemotion CC-BY-SA-4.0 + rare PMC CC-BY-SA) | 1,897 | Separate SA file — honour ShareAlike |
| empty_unknown (EMPTY / no EPMC hit) | 8,963 | **Excluded** from commercial Zenodo pool |
| other (CC-BY-ND via Crossref recovery) | 5 | Held aside — `licence_pools/irexp_other.jsonl.gz` |
Provenance (unchanged totals): PMC-sourced 119,345; Chemotion 1,888.
Unique PMCIDs joined: 15,416. See `docs/scientific_data/LICENCE_REMEDIATION.md`.
Materialise pools
-----------------
```
python scripts/join_pmc_licences.py # (re)join + stamp + write licence_pools/
python scripts/split_license_pools.py # report
python scripts/split_license_pools.py --write OUT/ # by license_pool
python scripts/split_license_pools.py --write-sources OUT/ # pmc vs chemotion only
```
Pre-built files live under `data/irexp/licence_pools/`.
Attribution
-----------
Re-users must (a) cite the dataset via its archival Zenodo DOI (see PAPER.md
"Data and code availability"), (b) attribute originating publications via each
record's `source_doi` / `pmcid`, and (c) honour each record's `license` /
`license_pool`. Commercial ML redistribution should use the **commercial** pool
only. Chemotion-derived or mixed-SA redistributions must remain CC-BY-SA-4.0.
Code
----
The code in this repository is released under the MIT License (see top-level LICENSE).
Hugging Face
------------
The public mirror `ilkhamfy/IRexp` previously overclaimed blanket CC-BY for PMC rows.
Re-issue after remediation — see `docs/scientific_data/LICENCE_REMEDIATION.md` § HF
re-upload. No credentials are committed here.
|