Data sources and licensing
Every fact in Radical Mandarin names where it came from: every reading, meaning, part, radical, band and stroke count carries the source it was taken from and how. Open any entry in the Explorer and its last section, “Source of this information”, lists them. This page is the table those notes link to: each source, its license, what the corpus takes from it, and what we owe it. Corrections and derivations are our own and say so. Corpus version 18; 0 facts without a source.
In the corpus
makemeahanzi (dictionary.txt)
LGPL v3 or later · in the corpus · T2
5,769 facts: meaning, parts, origin type, origin note, the entry, reading.
Character decompositions, etymology notes, and component roles from makemeahanzi (Shaunak Kishore), LGPL v3+.
What we owe: Ship the license text; keep makemeahanzi-derived fields a separable component in any published dataset.
complete-hsk-vocabulary (CC-CEDICT content)
MIT (compilation); contained CC-CEDICT content CC BY-SA 4.0 · in the corpus · T1 via T2
9,561 facts: meaning, reading, part of speech, traditional form.
Word forms assembled via complete-hsk-vocabulary (MIT).
What we owe: Keep the MIT notice; CC-CEDICT obligations below apply to the contained content.
CC-CEDICT (MDBG)
CC BY-SA 4.0 · in the corpus · T1
6,955 facts: meaning, reading, traditional form.
Word meanings, pinyin spellings, and traditional forms include CC-CEDICT content (MDBG), CC BY-SA 4.0.
What we owe: Attribute CC-CEDICT/MDBG; share-alike applies to the dictionary content we redistribute (meanings, pinyin spellings, traditional forms), which since 2026-09-16 includes the committed gate slice data/cedict-grammar-slice.txt (verbatim entries, built by scripts/build-cedict-slice.ts).
Pinned copy: https://www.mdbg.net/chinese/export/cedict/cedict_1_0_ts_utf-8_mdbg.txt.gz · taken 2026-09-16
Unicode Unihan database
Unicode License v3 · in the corpus · T0
4,515 facts: radical, stroke count, reading, the entry.
Character readings, definitions, and radical assignments from the Unicode Unihan database (Unicode License).
What we owe: Reproduce the Unicode copyright notice.
Pinned copy: https://www.unicode.org/Public/UCD/latest/ucd/Unihan.zip · version 18.0.0 · taken 2026-09-16
Unicode CJKRadicals.txt and EquivalentUnifiedIdeograph.txt
Unicode License v3 · in the corpus · T0
2,486 facts: radical, radical number, radical form, form, the entry.
Radical-number mapping and radical-glyph equivalences from the Unicode Character Database.
What we owe: Covered by the Unicode notice above.
Pinned copy: https://www.unicode.org/Public/UCD/latest/ucd/CJKRadicals.txt · taken 2026-09-16
English Wikipedia, "Kangxi radicals" (pinned revision 1362962863)
CC BY-SA 4.0 · in the corpus · T1 (community-reviewed; the page verified by a native reader, F4)
758 facts: radical name, radical reading, radical name (Chinese), radical form, radical number, form, radical, written-as.
Radical names, meanings, classroom names, variant and simplified forms, readings and stroke counts from Wikipedia's "Kangxi radicals" list (Wikipedia contributors, CC BY-SA 4.0), revision 1362962863.
What we owe: Attribute Wikipedia contributors with the revision link; share-alike on the derived radical table (data/staged/kangxi-radicals.json, radicals.json and what the app compiles from them); re-pin by revision id, never by rolling fetch (docs/source-of-truth.md 2.12).
Pinned copy: https://en.wikipedia.org/w/index.php?title=Kangxi_radicals&oldid=1362962863&action=raw · taken 2026-09-16
ivankra/hsk30 (verified 2021 transcription)
MIT · in the corpus · T0 via T2 (sample-verified)
4,907 facts: reading, the entry.
2021 standard levels and pinyin via the ivankra/hsk30 transcription (MIT), sample-verified against the standard's PDF.
What we owe: Keep the MIT notice.
Pinned copy: https://raw.githubusercontent.com/ivankra/hsk30/4ff9e3915/hsk30.csv · taken 2026-09-03
HSK exam syllabus 2026 (chinesetest.cn PDF)
Official syllabus document; facts (bands, readings) used, text not reproduced · in the corpus · T0
8,104 facts: band, the entry, writing band, reading.
2026 HSK band assignments verified against the official exam syllabus (chinesetest.cn).
What we owe: Cite the syllabus; do not reproduce the document.
Pinned copy: https://hsk.cn-bj.ufileos.com/3.0/%E6%96%B0%E7%89%88HSK%E8%80%83%E8%AF%95%E5%A4%A7%E7%BA%B21219.pdf · taken 2026-09-16
汉语拼音方案 (Scheme for the Chinese Phonetic Alphabet, 1958)
State document; the tables (initials, finals, spelling rules) are facts, reproduced as such · in the corpus · T0
829 facts: final, initial, the entry.
The initials, finals and pinyin spelling rules follow the 汉语拼音方案 (1958).
What we owe: Cite the scheme.
English Wikipedia, "Pinyin" and "Standard Chinese phonology" (read 2026-09-14)
CC BY-SA 4.0 (facts only: broad IPA transcriptions) · in the corpus · T1
57 facts: IPA.
IPA transcriptions of the initials and finals after Wikipedia's "Pinyin" and "Standard Chinese phonology" articles (Wikipedia contributors, CC BY-SA 4.0).
What we owe: Attribute Wikipedia contributors; share-alike on the derived sound tables.
The Unicode Standard (ideographic description characters U+2FF0..U+2FFB; radical-block names)
Unicode License v3 · in the corpus · T0
16 facts: the entry, written-as.
Character structure types are the Unicode Standard's ideographic description characters.
What we owe: Covered by the Unicode notice.
Radical Mandarin curation layer (data/curation)
CC BY-SA 4.0 on release (docs/licensing.md) · our own · T3 (reviewed corrections, each citing its authority)
31 facts: reading, the entry, meaning, part of speech, traditional form.
Corrections are Radical Mandarin's own, each citing the authority it follows.
What we owe: Each entry names its source, reason, date and author.
Radical Mandarin Proprietary Pipeline
CC BY-SA 4.0 on release (docs/licensing.md) · our own · T3 (rules stated in the scripts and the ontology principles)
9,441 facts: the entry, meaning, base, tone, final, initial, form, reading, radical form, radical number, first tone, second tone, examples.
Derived facts (memberships, bands of tables, forms in use, classes and shapes) are computed by the Radical Mandarin Proprietary Pipeline from the sources above.
What we owe: None.
Gaps, stated plainly
- Origin notes and part roles come from makemeahanzi, one author's account with no cited authority; a second source (Wiktionary glyph origins) is planned.
- Guides, grammar-pattern statements and examples, and the sound tables' classroom descriptions are written by the app's authors and await native-speaker review (the audit track).
- Audio clips are synthesized (Azure Neural TTS, voice zh-CN-XiaoxiaoNeural); the clip store records the voice and the spoken text, not the service version or the date of synthesis.
- Sentences and readings carry their own attribution on their own pages (Tatoeba contributors, public-domain works, Chinese Wikipedia excerpts); they are not counted in this table.
- Stroke counts follow Unicode 18's kTotalStrokes, which counts 阝 as three strokes and 艹 as four (the Kangxi convention); PRC dictionaries count 那 as six.
- Part-of-speech tags come from complete-hsk-vocabulary's machine taggers and stay hidden in the app until sourced.
- The 2021 standard's word list reaches the corpus through a community transcription (ivankra/hsk30), sample-verified against the standard's image-only PDF.
Checked against
English Wiktionary, "Appendix:Mandarin Frequency lists" 1-5000 (pinned revisions)
CC BY-SA 4.0 · checked against · T2 (a CKIP / Academia Sinica list as transcribed by Wiktionary)
What we owe: Analysis only (the vocabulary-gap instrument); attribution and share-alike if any rank or membership fact is ever ingested.
Pinned copy: https://en.wiktionary.org/w/index.php?title=Appendix:Mandarin_Frequency_lists/1-1000&oldid=91542360&action=raw · taken 2026-09-15
GF0025-2021 standard (Ministry of Education PDF)
Official standard document; facts (membership, levels, readings) used, text not reproduced · checked against · T0
4,907 facts: reading, the entry.
HSK 3.0 level membership and readings verified against GF0025-2021.
What we owe: Cite the standard; do not reproduce the document.
Pinned copy: http://www.moe.gov.cn/jyb_xwfb/gzdt_gzdt/s5987/202103/W020210329527301787356.pdf · taken 2026-09-03
elkmovie/hsk30 (OCR word + char lists)
MIT (Pleco Inc.) · checked against · T2
What we owe: Keep the copyright notice if redistributed.
Pinned copy: https://raw.githubusercontent.com/elkmovie/hsk30/7f3d4fd/wordlist.txt · taken 2026-09-03
hermitdave FrequencyWords (zh_cn)
MIT (code) / Creative Commons (lists) · checked against · T2
What we owe: Attribution if ingested; cross-check only for now.
Pinned copy: https://raw.githubusercontent.com/hermitdave/FrequencyWords/525f9b560de45753a5ea01069454e72e9aa541c6/content/2018/zh_cn/zh_cn_50k.txt · taken 2026-09-03
CHISE / cjkvi-ids decomposition database
GPL-family (CHISE terms) · checked against, never copied · T2
What we owe: Checker only: compared against, never ingested or redistributed, so no GPL obligations attach to our data.
Pinned copy: https://raw.githubusercontent.com/cjkvi/cjkvi-ids/master/ids.txt · taken 2026-09-16
Consulted or under assessment
- wordfreq (frequency data) (Apache-2.0 (code); data CC BY-SA 4.0 with recorded SUBTLEX permission; under assessment)
- English Wiktionary Chinese extract (kaikki/wiktextract) (CC BY-SA (dual GFDL); we take the CC BY-SA grant; under assessment)
- Tatoeba sentence corpus (CC BY 2.0 FR (per-sentence attribution); CC0 subset; under assessment)
- Universal Dependencies Chinese GSDSimp (CC BY-SA 4.0; under assessment)
- Public-domain modern vernacular shelf (pre-1929 works, authors d. before 1976) (Public domain (China life+50; US pre-1929 publication); under assessment)
- Chinese Wikipedia (excerpts) (CC BY-SA 4.0; under assessment)
- Chinese Grammar Wiki (and the UD learner treebank from it) (CC BY-NC-SA (non-commercial); consulted only)
- zhongwen.com (Harbaugh genealogies) (Copyrighted, no reuse license; consulted only)
- Taiwan MoE dictionaries (moedict) (CC BY-ND (no derivatives), verify per dataset; consulted only)
- Shuowen Jiezi (Xu Shen, ~100 CE) (Public domain; consulted only)