Skip to content

LICENSE-DATA.md

Latest commit

 

History

History
48 lines (35 loc) · 2.41 KB

LICENSE-DATA.md

File metadata and controls

48 lines (35 loc) · 2.41 KB
Raw

Chem2TextQA dataset license and upstream terms

Dataset release

Chem2TextQA, the derivative dataset produced by this pipeline, is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).

Upstream sources and their terms

Chem2TextQA derives from three publicly redistributable sources, whose terms continue to apply transitively to any user of this dataset:

Upstream Terms
PubChem (NIH/NLM) Public domain (no copyright claimed by NIH/NLM). Reuse requires attribution but no license compatibility is enforced. https://www.ncbi.nlm.nih.gov/home/about/policies/
PubMed (abstracts via NIH/NLM) No copyright claimed on the bibliographic database; individual abstract copyright is held by the original publishers. Use in research and aggregate is permitted; full redistribution of abstract text may be restricted.
PMC Open Access Subset Articles carry their original author-selected licenses (CC BY, CC BY-NC, CC BY-SA variants). Chem2TextQA redacts and extracts sentence fragments; derivative-work status depends on fragment lengths and jurisdiction.

Important reservations

  1. Sentence fragments in evidence bundles. Chem2TextQA's Phase 0 evidence sentences are extracted verbatim from PubMed abstracts and PMC full-text articles (with compound names replaced by [COMPOUND]). Users redistributing the raw evidence bundles must respect the original publisher / author licenses. Where PMC articles are licensed CC BY-NC, derivative non-commercial use is required.
  2. Q&A pairs. Phase 1-3 LLM outputs are model-generated derivative text inspired by the upstream evidence. These are released under CC BY 4.0 for Chem2TextQA as a whole, but users should be aware that they encode (in paraphrased form) content from the upstream sources.
  3. SMILES / structural metadata. PubChem-sourced; public domain.

Recommended attribution

Citations as a user of the dataset:

Chem2TextQA: A SMILES-Grounded Q&A Dataset Derived from PubChem × PubMed × PMC. [Authors anonymized for review.] 2026.

Include citations to PubChem, PubMed, and any specific PMC articles surfaced via the PMID metadata preserved in each compound record.

If you are a rights-holder with a concern

See the dataset contact in the Hugging Face dataset card (post-release). Chem2TextQA will honor reasonable takedown / correction requests for specific evidence records.