Data Copyright & Sources
No rights claimed over the works; per-track source licences; sources and removal logs
1. Our position
midicn-lib aggregates other people's work, and we claim no rights over the works in it.
This differs fundamentally from the knowledge sites, where we hold the copyright and can grant CC BY 4.0:
| Knowledge sites | These sites (data) | |
|---|---|---|
| Content | Text and diagrams we wrote/drew | Works released by others under open licences |
| Can we license it? | ✅ Yes | ❌ No |
| Licence source | Our CC BY 4.0 grant | Each work's own source licence |
| Commercial use | Possible (incl. advertising) | Non-commercial |
2. Where the licences come from
Every source in the dataset is published under an explicit open licence. Main categories:
| Licence | Meaning | Typical sources |
|---|---|---|
| Public domain | Copyright expired or waived | Transcriptions of old classical and traditional material |
| CC0 | All rights waived by the author | Some modern contributors |
| CC BY | Attribution only | Various academic and hobbyist projects |
| CC BY-SA | Attribution + share-alike | Several score communities |
| CC BY-NC-SA | Non-commercial + share-alike | A particular piano dataset |
| OPEN / custom open terms | Defined per project | Several folk archives |
| Traditional-music transcriptions | Study and research only | Folk-song collections |
The specific licence is stated per track on each page and can be checked in the dataset's licence ledger.
3. Our compliance work
"Aggregating open content" warrants care. Here is what we do:
1. Per-source licence verification
Before inclusion, each source is checked on four points: licence type · redistribution permitted · attribution required · commercial restrictions. Sources that do not permit open redistribution are not included.
2. Removal of possibly in-copyright works
For material in one source that appeared to be modern popular/commercial music, we ran a dedicated review, removed it, and kept a line-by-line log (source, title, reason). Detection relies on directory-name and file-path features — see the dataset's internal review records.
3. Per-track labelling and traceability
- Every record carries a source marker and a licence marker
- The dataset's sources and changes are recorded
- Metadata follows "leave blank rather than guess" — uncertain values stay empty
4. Original licences unchanged
We do not relicense these works, add no extra restrictions, and claim no new rights arising from the curation.
4. If you believe a work's rights are infringed
We will act promptly. Such claims matter especially here, because we cannot vouch for the provenance of every work.
Please provide:
1. The specific work (page URL or dataset identifier) 2. The work you claim, and the nature of your rights in it 3. How to reach you 4. A short explanation of why inclusion exceeds its licence
Submit via an issue on our GitHub repository.
Our commitments:
- On a valid notice, we take it down first and verify after
- If the claim proves mistaken, we explain why and restore it
- We will not ask for more information than necessary
5. About the curation itself
While we claim no copyright in the works, the curation (structured metadata, de-duplication, source labelling, the zoning system) is work we put in.
We place no extra restrictions on it — you may use and redistribute the structured data freely, and we ask only that you note it came from midicn-lib. That is a request, not a condition.
This page is a plain-language summary, not legal advice.
Sources
- 本站自订声明，据实描述数据集的来源与授权状况
- midicn-lib 逐来源许可台账
Used for fact-checking only; all prose is original.