造序注音 ZaoSeq Bopomofo — public core
Copyright 2026 ZaoSeq Technologies (造序科技，籌備中)

The source code in this repository (src/, tests/) is licensed under the Apache License, Version 2.0.
See LICENSE.

Data in data/ and benchmarks/ is NOT covered by the Apache License unless stated otherwise:

- data/cns11643/
  Derived from CNS11643 中文標準交換碼全字庫, 數位發展部 (Ministry of Digital Affairs, Taiwan),
  used under the Open Government Data License, version 1.0 (政府資料開放授權條款-第1版),
  https://data.gov.tw/license . Source: https://data.gov.tw/dataset/5961

- data/corpus/
  Manifests and statistics derived from Taiwan government open datasets, each under the license recorded
  for that source in data/sources.json and data/SOURCES.md (Open Government Data License, version 1.0).

- data/daily/
  Source audit records, manifests and statistics for daily-language corpora. No third-party corpus text is included.
  Tatoeba (https://tatoeba.org): CC BY 2.0 FR; data/daily/attribution/ lists the authors that must be credited.
  Mozilla Common Voice Scripted Speech 27.0 - Chinese (Taiwan), obtained from Mozilla Data Collective
  (https://mozilladatacollective.com/datasets/cmu5wkw5i00c6o107jnm3t51u): CC0 1.0. Only sentence ids and sentence
  text were used, never speaker data. The platform terms do not allow re-hosting, so no Common Voice data is included.
  Unicode Unihan database (Unicode License v3) is used only to detect simplified characters; it is not included.

- benchmarks/test/gov.jsonl, benchmarks/test_v1/gov.jsonl
  Clauses taken from Taiwan government open datasets, under the license recorded for each source in data/sources.json
  (Open Government Data License, version 1.0).

- data/builtin/ and the other files in benchmarks/ (evaluation sets written for this project and results)
  Written by the ZaoSeq Bopomofo project and licensed under the Apache License, Version 2.0.

This distribution does not contain any model weights. The production contextual ranker, its training code
and its model artifacts are not part of this repository and are not released under the Apache License.
