# Third-Party Notices This project incorporates code and data derived from external sources. Attribution and license obligations for each are listed below. The project itself — code and dictionary data alike — is licensed under the Apache License 2.0 (`LICENSE.Apache`), except where a section below states otherwise. --- ## sudachi.rs The morphological analyzer in `src/analyzer/` is derived from [sudachi.rs](https://github.com/WorksApplications/sudachi.rs), the Rust implementation of the Sudachi Japanese morphological analyzer, ported to TypeScript. - **Project**: - **Rightsholder**: Works Applications Co., Ltd. - **License**: Apache License 2.0 — . - **Local notice**: `LICENSE.Apache`. - **Modifications**: ported from Rust to TypeScript and adapted to the project's analyzer interface. Dictionary loading and plugin configuration are project-specific; the plugin algorithms themselves are ported. Files that contain ported code carry a header naming the upstream source file; files in `src/analyzer/` without that header were written for this project. The analyzer's default settings are converted from the default resources of Sudachi (`sudachi.json`, `char.def`, `unk.def`, `rewrite.def`). --- ## SudachiDict The binary system dictionary used by the analyzer (`system.hrdic`, `system.hridx`) is converted from the SudachiDict full dictionary (release 20260116) by a conversion tool that is not part of this package. The conjugation templates (`data/dict/analyzer-lexicon/conjugation-templates.json`) are generated from the same dictionary. - **Project**: - **Rightsholder**: Works Applications Co., Ltd. - **License**: Apache License 2.0 — . - **Upstream resources**: SudachiDict incorporates UniDic (BSD-style license, The UniDic Consortium) and mecab-unidic-NEologd (Apache License 2.0, Toshinori Sato) together with the further sources acknowledged in its legal notice. - **Local notice**: `licenses/SudachiDict-LEGAL.txt` (SudachiDict `LEGAL`, verbatim — includes the UniDic redistribution conditions and disclaimer). - **Modifications**: the dictionary is re-encoded into the project's own binary format (lexicon and index files); entry content is not altered. --- ## Japanese WordNet Part of the morpheme dictionary entries in `data/dict/morpheme/` were built from Japanese WordNet synsets: Japanese headwords and their sense grouping are taken from Japanese WordNet, while the Korean renderings were produced independently for this project. - **Project**: - **Rightsholders**: National Institute of Information and Communications Technology (NICT); Francis Bond; Takayuki Kuribayashi. - **License**: Japanese WordNet license (permissive; the copyright notice, statements and disclaimer must accompany all copies, including modifications). - **Local notice**: `licenses/japanese-wordnet.txt` (verbatim). - **Modifications**: synset lemmas are extracted, filtered and merged into the project's morpheme dictionary schema with project-authored Korean translations. --- ## KANJIDIC2 The kanji reading tables in `src/third-party/kanjidic2/` are derived from the KANJIDIC2 file. They are consumed at runtime by the kanji-reading alignment used for long-vowel boundary detection in Japanese-to-Korean transliteration and by the Sino-Korean reading generator. They are kept in their own directory so that their independent license terms remain clear. - **Project**: - **Rightsholder**: Electronic Dictionary Research and Development Group (EDRDG), Monash University, Australia. - **License**: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) — . This license applies to the files in `src/third-party/kanjidic2/` only. - **Local notice**: `src/third-party/kanjidic2/LICENSE`. - **Modifications**: only ``, the base portion of `` (okurigana `.` tail and prefix/suffix `-` markers stripped), `` and `` readings are extracted; Japanese readings are normalized to katakana; entries are serialized as TypeScript object literals. EDRDG's attribution requirements are met by this file, the local `LICENSE` file and the attribution headers in the generated data modules. Downstream redistributors must preserve all three. --- ## Unihan (kHangul, kKorean) `src/third-party/unihan/hanja-readings.ts`, a table of representative Korean readings for CJK unified ideographs not covered by KANJIDIC2, is generated from the Unihan database. - **Source**: Unicode Character Database — Unihan (). - **Rightsholder**: Unicode, Inc. - **License**: Unicode License v3 — . - **Local notice**: `src/third-party/unihan/LICENSE` (verbatim). - **Modifications**: one representative reading per character is selected from `kHangul`/`kKorean` and serialized as a TypeScript object literal.