# Methodology How a license status is decided and checked. Source: schema/README.md. [Quality, in numbers Excerpt checks, review, audits, known issues](https://mdexa.com/quality/) ## Principle Read restrictions broadly, permissions strictly. A code repository license is not a data license. ## The five rules 1 Counts as a data license | The license field of the dataset's hosting page (Hugging Face card, Zenodo, Figshare); a license or use statement in the docs aimed at the data; the LICENSE of a repository that holds only data. | 2 Repository license only | The LICENSE of a repository that mixes code and data, when the docs do not say it covers the data. | 3 Repository license, no data license | license_status is unstated, hosting is pending; the repository license is kept in repo_license. | 4 Badge only | A README badge whose LICENSE file is missing from the snapshot: repo_license.source is badge-only, not verified. | 5 Second-hand evidence | Entries without a URL, or seen only in search snippets or mirrors, are marked secondhand and never decide permissive or restricted alone. | ## Axis 1 · license_status: what the source says Permissive | `permissive` | Every statement agrees and allows commercial use and redistribution. | 44 | Restricted | `restricted` | A statement forbids commercial use or adaptation, limits use to research, or the rights belong to a third party or require payment. | 29 | Conflict | `conflict` | Statements in different places contradict each other. | 4 | Unstated | `unstated` | No data license statement was found. | 14 | Unverifiable | `unverifiable` | The page could not be opened, or only second-hand information was found. | 10 | ## Axis 2 · hosting: what Mdexa does Hosted | `hosted` | Mdexa serves the files. Needs a permissive license, no risk flags and a human sign-off. | 0 | Eligible | `eligible` | Permissive with no risk flags, waiting for a human sign-off. | 0 | Link only | `link-only` | Indexed and linked to the source: the license is restricted, or permissive with a risk flag. | 73 | Pending | `pending` | License in conflict, unstated or unverifiable; linked to the source until resolved. | 28 | ## Evidence and checks - **Fetch** Each source page is saved as a text snapshot with its sha256 and fetch time: 1098 snapshots. - **Excerpt** License statements are copied verbatim with their location: 426 evidence entries. - **Check** A script finds every excerpt in its snapshot: 401 / 402 found. - **Decide** The 26 rules above set license_status; hosting follows from it, the risk flags and a human sign-off. - **Gate** 5 machine checks per card (validate, hosting, numbers, links, quotes): 97 pass, 4 to review, 0 fail. - **Publish** 101 cards as page, JSON and Markdown, with last_verified on each. ## Public standards ICD-11 | Disease classification reference; disease links switch on after physician review | GB/T 15657-2021 | Classification and codes of TCM diseases and patterns | SPDX | License identifiers on every card | schema.org Dataset | Machine-readable metadata on every card page | ## Data scope - Datasets are organised by system (Modern medicine, Traditional medicine, Cross-cutting). - No patient records are hosted. - Third-party datasets derived from clinical text are flagged: 15 today. - Not legal advice: a status records what the license text says. ## Rights holders If a card misstates a license, write to us. [Rights holders](https://mdexa.com/takedown/) ## Report [License check summary](https://mdexa.com/reports/license-audit/) Draft, to be published after checking with the authors. The data, field spec and license rules are public on [GitHub](https://github.com/mdexa-ai); dataset authors can file an issue to correct a card.