# HSK character and vocabulary data sources

The current distribution candidate includes level-specific written forms,
pinyin, original parts of speech, and manually reviewed example words and
sentences for all HSK levels. It also includes manually reviewed English and
Korean meanings and sentence translations. It does not include example-sentence
images. The following authoritative materials were used to review character and
word coverage:

- Chinese Testing International, `HSK-2015.xlsx`: the HSK 2.0 level-specific
  vocabulary list, from which level-specific word forms and character coverage
  were derived.
- Center for Language Education and Cooperation, *Chinese Proficiency Test HSK
  Syllabus* (published November 2025, effective July 2026): level-specific word
  forms, pinyin, parts of speech, and the 3,088-character recognition list.
- Ministry of Education and State Language Commission, GF 0023-2020,
  *Standard of Stroke Orders of the Commonly Used Standard Chinese Characters*:
  the 8,105 mainland-standard characters and their stroke counts.
- Chinese Testing International's query system for the *Chinese Proficiency
  Grading Standards for International Chinese Language Education*: pinyin
  cross-checking.
- *Scheme for the Chinese Phonetic Alphabet*, approved by the National People's
  Congress and published by the Ministry of Education: initial ordering and
  tone notation.
- GB/T 16159-2012, *Basic Rules of the Chinese Phonetic Alphabet Orthography*:
  sentence pinyin and pedagogical tone-change notation for `一` and `不`.
- State Council / Ministry of Education, *Table of General Standard Chinese
  Characters*: standard-character numbering and tier boundaries.

Stable URLs and SHA-256 hashes of downloaded source materials are recorded in
`data/hsk/source/metadata.json` in the KyouStudy source repository. Source
changes must update both the content version and pinned source hashes.

KyouStudy does not redistribute the official XLSX or PDF source files. It stores
only normalized written forms, pinyin, parts of speech, and other derived data
needed for level assignment in JSON and SQLite form. Homograph numbering,
layout, and other source-reproduction information are not retained.

Temporary sentences from `hsk_app_data` were used only as candidates. Every
final example was reviewed again for character and word relationships,
mainland-standard Simplified Chinese, pinyin, tones, neutral tone, erhua, and
sentence naturalness. Only the final level-specific example JSON and its pinned
review manifest are generation inputs; `hsk_app_data` is not a runtime or build
dependency.

Stroke outlines and medians for all HSK characters were extracted from
`hanzi-writer-data` 2.0.1. The data derives from Make Me a Hanzi and is licensed
under the Arphic Public License. It is separate from the MIT-licensed Hanzi
Writer runtime code and from the Japanese KanjiVG assets. See
[`HANZI_WRITER_DATA.md`](HANZI_WRITER_DATA.md) and
[`ARPHICPL.TXT`](ARPHICPL.TXT).