# Data license

The generated corpus is distributed under the
[Creative Commons Attribution-ShareAlike 4.0 International license](https://creativecommons.org/licenses/by-sa/4.0/)
(`CC BY-SA 4.0`).

The Wikipedia portion is derived from Japanese Wikipedia and remains subject
to its applicable attribution and share-alike terms. Each record retains a URL
to the corresponding article so that its contributors and revision history can
be identified.

The Aozora Bunko portion is restricted by this pipeline to works for which both
the work and every associated person have the official copyright flag `なし`.
Those source texts are public-domain works. Bibliographic metadata and source
URLs are retained as requested by Aozora Bunko's file-handling guidelines.

This notice does not replace the terms attached to either upstream source. See
`NOTICE.md` for source details.

The generated Mozc-format n-gram dictionaries are derived from the same corpus
and are distributed under CC BY-SA 4.0. The bundled Mozc context-ID definition
and third-party notice files retain their respective upstream terms; see the
Release assets `MOZC-LICENSE`, `IPADIC-COPYING`, and `IPADIC-NOTICE`.

The hiragana-to-English dictionary is derived from the Electronic Dictionary
Research and Development Group's JMdict file and is distributed under the
same CC BY-SA 4.0 terms. The exact source snapshot and the EDRDG licence page
are included in each Release as `JMdict_e-<date>.xml.gz` and
`JMDICT-LICENSE.html`. This project does not claim copyright over JMdict
material.

The AJIMEE-Bench evaluation dataset is not included in JapaneseCorpus
Releases. Its aggregate accuracy report contains provenance and factual metric
values only; the upstream dataset remains subject to its CC BY-SA 3.0 terms as
described in `NOTICE.md`.