v2026.404.0 - 2026-04-04
v2026.322.0 - 2026-03-22
- romkan: "she"/"sshe" ローマ字マッピングを追加 by @tokuhirom in #547
v2026.318.0 - 2026-03-18
- romkan: "she"/"sshe" ローマ字マッピングを追加 by @tokuhirom in #547
v2026.313.0 - 2026-03-13
- ユーザー辞書の自動学習データ分離: 課題整理と実装方針 by @tokuhirom in #544
- romkan/keymap の設定ファイルを YAML から JSON に移行 by @tokuhirom in #543
- compound_dict の自動学習データを SKK-JISYO.user から分離 by @tokuhirom in #546
v2026.312.0 - 2026-03-12
- Add la, li, lu, le, lo, ltsu, ltu, lya, lyu, lyo, lwa by @mattn in #540
- romkan/keymap YAML に name/description を追加 by @mattn in #542
v2026.310.2 - 2026-03-10
- 半角カタカナを変換候補に追加 by @tokuhirom in #538
v2026.310.1 - 2026-03-10
v2026.310.0 - 2026-03-10
v2026.303.0 - 2026-03-03
- 「各項目」「各プロジェクト」の辞書複合語登録 by @tokuhirom in #527
- LatticeGraph に unknown_bigram_cost フィールドを追加 + 辞書複合語登録 by @tokuhirom in #529
- ユーザー bigram 学習に BOS/EOS を追加 by @tokuhirom in #530
v2026.227.0 - 2026-02-27
- 「うつりこんでる」の分節崩壊を修正し、tokenize-line.sh を改善 by @tokuhirom in #521
- 辞書に「開ける/ひらける」「サイドバー」を追加 by @tokuhirom in #523
- 1語候補の未知bigramペナルティ緩和をリランキングに追加 by @tokuhirom in #524
- 「かいてない」→「買いてない」の誤変換を修正 by @tokuhirom in #525
v2026.225.2 - 2026-02-25
- perf: rsmarisa Agent を再利用して trie 検索のアロケーションを削減 by @tokuhirom in #518
- perf: bigram/skip-bigram の key/value 分離で lookup 高速化 by @tokuhirom in #520
v2026.225.1 - 2026-02-25
- corpus-stats: Fix normalization examples in README by @omasanori in #515
- refactor: must.txt のエントリを should.txt に移動 by @tokuhirom in #517
- 辞書: 「誤検知」を追加 by @tokuhirom in #514
v2026.225.0 - 2026-02-25
- fix: 空コーパスで evaluate が panic する問題を修正 by @tokuhirom in #504
- fix: 「ろぐ」→「ログ」の変換を修正 by @tokuhirom in #506
- fix: FFI コールバックの null ポインタチェックを追加 by @tokuhirom in #507
- fix: f32 ソートの NaN panic 防止と config.h 自動生成 by @tokuhirom in #509
- fix: 「すき」→「数寄」のBOS bigram過剰学習を修正 by @tokuhirom in #508
- fix: tokenize の並列処理でエラー集約し即 panic を防止 by @tokuhirom in #510
- perf: learn-corpus の高速化 — 収束済みの文をスキップ by @tokuhirom in #511
- perf: HashMap を FxHashMap (rustc-hash) に統一 by @tokuhirom in #512
- Sudachi 辞書からカタカナ普通名詞を取り込み by @tokuhirom in #513
v2026.224.0 - 2026-02-24
- fix: 「したい」→「肢体」の誤変換を修正 by @tokuhirom in #499
- ci: apt パッケージキャッシュで CI を高速化 by @tokuhirom in #501
- fix: 「さい」→「賽」の誤変換を修正 by @tokuhirom in #502
v2026.220.3 - 2026-02-20
- 「他責」を辞書・コーパスに追加 by @tokuhirom in #494
- 「あらいだそう」→「洗い出そう」の変換を修正 by @tokuhirom in #496
- feat: Sudachi 辞書から固有名詞を SKK-JISYO.akaza に統合 by @tokuhirom in #497
- fix: 辞書エントリのクリーンアップフィルタを追加 by @tokuhirom in #498
v2026.220.2 - 2026-02-20
- Add pub lookup methods to unigram and bigram LM by @tokuhirom in #492
v2026.220.1 - 2026-02-20
- feat: tagpr リリース時に Cargo.toml のバージョンを自動更新 by @tokuhirom in #489
v2026.220.0 - 2026-02-20
- feat: モデルファイルにバージョン情報を埋め込む by @tokuhirom in #487
v2026.218.1 - 2026-02-18
- fix: install-model のダウンロード元をモノレポに修正しタグを固定 by @tokuhirom in #481
- fix: 助数詞単漢字の LM スコアフォールバックを追加 by @tokuhirom in #483
- feat: 「誤情報」の辞書登録 + チューニング知見・check使い方追記 + vibrato辞書ターゲット追加 by @tokuhirom in #484
- feat: 「誤情報」の辞書登録・コーパス追加 + vibrato辞書ターゲット追加 by @tokuhirom in #485
- fix: 「誤情報」コーパスを複合語登録に修正（ご→誤 退行修正） by @tokuhirom in #486
v2026.218.0 - 2026-02-18
- docs: 評価方法のドキュメントを追加 by @tokuhirom in #454
- fix: convert メソッドでリランキングが適用されない問題を修正 by @tokuhirom in #455
- feat: 数字+接尾辞の複合トークン正規化改善 by @tokuhirom in #457
- refactor: KANSUJI 動的マーカーを廃止し静的候補生成に統一 by @tokuhirom in #458
- akaza-default-model と akaza-corpus-stats をモノレポに統合 by @tokuhirom in #459
- fix: UniDic CSJ ダウンロードを GitHub Release ミラーに変更 by @tokuhirom in #460
- docs: README 重複整理（リポジトリ統合後） by @tokuhirom in #461
- docs: corpus-stats/docs/ を docs/ に移動 by @tokuhirom in #462
- docs: default-model/docs/ を docs/ に移動 by @tokuhirom in #463
- docs: UniDic カタカナ語抽出の説明を data-flow.md に追加 by @tokuhirom in #464
- 実装完了: 数値+助数詞変換を再設計して動的候補生成を拡張 by @tokuhirom in #465
- 百の位の促音便形をかな数詞パーサに追加 by @tokuhirom in #467
- fix: corpus-stats release 先を akaza-im/akaza-corpus-stats に修正 by @tokuhirom in #468
- chore: corpus-stats v2026.0216.0 に更新 + 評価履歴の参照先を追記 by @tokuhirom in #469
- feat: evaluate 結果のファイル保存改善（ブランチ名・corpus-stats バージョン追加） by @tokuhirom in #470
- docs: corpus-stats v2026.0216.0 比較レポート by @tokuhirom in #471
- BAD 傾向分析スクリプトと分析レポートの追加 by @tokuhirom in #472
- fix: 小文字かな始まりの不正辞書エントリの混入を防止 by @tokuhirom in #473
- docs: 四字熟語ソース調査結果をBAD傾向分析レポートに追記 by @tokuhirom in #474
- fix: 数値複合語の LM スコア選択でリテラルと正規化版の良い方を採用 by @tokuhirom in #475
- chore: tagpr を CalVer モードに変更 by @tokuhirom in #477
- fix: 「失敗しないように」が「失敗し内容に」になる分節崩壊を修正 by @tokuhirom in #476
- chore: tagpr を v1.17.0 に更新 by @tokuhirom in #479
v0.5.0 - 2026-02-12
- docs: k-best リランキング機構の設計方針 by @tokuhirom in #430
- feat: k-best リランキング機構を実装 by @tokuhirom in #432
- fix: evaluate/check でリランキングが反映されるように修正 by @tokuhirom in #433
- docs: リランキング評価レポートを追加 by @tokuhirom in #434
- feat: length-weight のデフォルト値を 2.0 に設定 by @tokuhirom in #435
- feat: skip-bigram リランキングの実装 by @tokuhirom in #436
- feat: skip-bigram を Viterbi DP に統合し変換精度を向上 by @tokuhirom in #437
- feat: learn-corpus で skip-bigram の学習に対応 by @tokuhirom in #439
- feat: lefthook による pre-commit フック設定を追加 by @tokuhirom in #438
- fix: check/evaluate の skip-bigram-weight デフォルトを 0.2 に修正 by @tokuhirom in #440
- feat: ユーザー適応学習に skip-bigram を追加 by @tokuhirom in #441
- perf: with_capacity 追加とバッファ再利用でメモリアロケーション削減 by @tokuhirom in #442
- perf: libakaza の HashMap/HashSet を FxHashMap/FxHashSet に置き換え by @tokuhirom in #443
v0.4.0 - 2026-02-11
- feat: 数字トークンの 正規化による LM スコア集約 by @tokuhirom in #426
- fix: 裸の数字トークンを 正規化から除外 by @tokuhirom in #428
- docs: 数字トークンの 正規化について README に追記 by @tokuhirom in #429
v0.3.0 - 2026-02-10
- akaza-data check コマンドを改善 by @tokuhirom in #380
- update deps by @tokuhirom in #383
- wordcnt-unigram: homograph_hack / score_hack を除去 by @tokuhirom in #385
- get_edge_cost のパフォーマンス改善 by @tokuhirom in #388
- add tokenize-line subcommand by @tokuhirom in #390
- use rsmarisa-0.3.0 for performance by @tokuhirom in #389
- learn-corpus コマンドのコーパスファイル説明を README に追加 by @tokuhirom in #391
- evaluate: [BAD] ログ行に入力の読みを追加 by @tokuhirom in #392
- tokenize-line: stdin からの複数行入力に対応 by @tokuhirom in #393
- evaluate: [BAD] ログを2行形式に簡潔化 by @tokuhirom in #394
- check: stdin からの複数行入力に対応 by @tokuhirom in #395
- README: かな漢字変換の仕組み・現状の制約・改善案を追記 by @tokuhirom in #396
- segmenter: unwrap() を除去し空文字列でのクラッシュを防止 by @tokuhirom in #397
- akaza-data tokenize のメモリアロケーション最適化 by @tokuhirom in #400
- k-best ビタビを導入し Tab キーで分節パターンを切り替え可能に by @tokuhirom in #398
- check: --k-best オプションで分節パターンを確認可能に by @tokuhirom in #401
- k-best 分節パターンとサジェスト機能の導入 by @tokuhirom in #403
- suggest 設定フラグを削除し、サジェストを常時有効化 by @tokuhirom in #404
- rsmarisa を 0.4.0 に更新 by @tokuhirom in #405
- README: 数字の動的変換の仕組みを追記 by @tokuhirom in #406
- BOS/EOS bigram を導入し、文頭・文末の文脈情報を活用 by @tokuhirom in #407
- akaza-data tokenize のメモリ確保を削減し高速化 by @tokuhirom in #408
- wfreq のメモリ確保を削減し高速化 by @tokuhirom in #410
- wordcnt-bigram の count_bigram を Vec 排除・1パス化で高速化 by @tokuhirom in #411
- wfreq/wordcnt-bigram の HashMap を FxHashMap に切り替え by @tokuhirom in #412
- evaluate: k-best スコアリングで BAD の重大度を分類 by @tokuhirom in #409
- 変換パイプラインのアロケーション削減 by @tokuhirom in #413
- 変換パイプラインの自明なアロケーション削減 by @tokuhirom in #414
- evaluate: 全角数字を半角に正規化してから変換する by @tokuhirom in #415
- evaluate: std::thread::scope でコーパス評価を並列化 by @tokuhirom in #416
- wfreq: fold+reduce で逐次マージし OOM を防止 by @tokuhirom in #417
- wfreq: redb でオンディスク集計し OOM を防止 by @tokuhirom in #418
- wfreq: バッチ commit でトランザクションオーバーヘッドを削減 by @tokuhirom in #419
- wfreq: 制御文字を含むワードのスキップ + parse_wfreq 堅牢化 by @tokuhirom in #420
- wordcnt-bigram: redb でオンディスク集計し OOM を防止 by @tokuhirom in #421
- vocab: wfreq パース時の unwrap panic を修正 by @tokuhirom in #422
- vocab: 日本語文字を含まない surface をフィルタ by @tokuhirom in #423
- k-best コスト表示バグ修正: 累積コストの二重カウントを解消 by @tokuhirom in #424
- feat: コーパスソースの重み付け対応 (path:weight 形式) by @tokuhirom in #425
v0.2.3 - 2026-02-05
- Guard crash risks and stabilize cache tests by @tokuhirom in #361
- AGENTS.md に cargo fmt 実行の注意を追記 by @tokuhirom in #362
- akaza-dict の引数不足クラッシュを回避 by @tokuhirom in #364
- make install で akaza-conf/akaza-dict もインストール by @tokuhirom in #365
- 設定保存時にconfigディレクトリを作成 by @tokuhirom in #366
- cargo-deny-action をCIで利用 by @tokuhirom in #367
- akaza-conf のOK後にウィンドウを閉じる by @tokuhirom in #368
- Shift+矢印で文節フォーカスを維持 by @tokuhirom in #369
- 入力モードメニューのチェック状態を同期 by @tokuhirom in #371
- README に ibus restart の記載を追加 by @tokuhirom in #370
- Clippy 警告の解消 by @tokuhirom in #373
- デバッグ用スクリプト名を整理 by @tokuhirom in #375
- ユーザー辞書メニューの初期表示と作成 by @tokuhirom in #372
- 設定メニューが動作しない問題を修正 by @tokuhirom in #376
- preedit テキストが黒背景に黒文字で見えない問題を修正 by @tokuhirom in #377
- Bump time from 0.3.46 to 0.3.47 by @dependabot[bot] in #378
v0.2.2 - 2026-02-03
- Migrate from marisa-sys to rsmarisa by @tokuhirom in #343
- Remove marisa-sys directory by @tokuhirom in #345
- Add comprehensive test coverage for conversion engine by @tokuhirom in #346
- Add CLAUDE.md with development guidelines by @tokuhirom in #347
- Fix crash risk with invalid keyval in ibus-akaza by @tokuhirom in #349
- Fix crash risk in extend_clause with empty yomi by @tokuhirom in #348
- Fix panic in build_string with out-of-bounds node_selected by @tokuhirom in #350
- CLAUDE.md の更新: PR 言語ポリシーとテストガイドラインの追加 by @tokuhirom in #352
- Phase 1: Convert ibus-akaza to lib+bin structure for testing by @tokuhirom in #356
- make install-model ターゲットを追加し、make install に統合 by @tokuhirom in #357
- gnome-terminal で日本語入力時にクラッシュする問題を修正 by @tokuhirom in #358
- Fix crash risks in graph_resolver with proper error handling by @tokuhirom in #351
- プリエディット属性の範囲指定ミスによる gnome-terminal クラッシュを修正 by @tokuhirom in #360
- ibus-akaza の unwrap/panic を穏当なエラーハンドリングに置換 by @tokuhirom in #359
v0.2.1 - 2026-01-27
- fix off by one for lookup table selection. by @tokuhirom in #257
- Refactor _update_lookup_table by @tokuhirom in #261
- ライブコンバージョンのルックアップテーブル処理を変更 Close #259 by @tokuhirom in #262
- kk /株式会社/ のようなアルファベットのみのものを辞書から読まないように by @tokuhirom in #264
- 数字を分節として取り出すようにする Close #263 by @tokuhirom in #265
- Optmize cargo deny by @tokuhirom in #266
- 漢数字入力機能 Close #67 by @tokuhirom in #267
- Refactor ibus by @tokuhirom in #269
- Use "default" layout. Close #244 by @tokuhirom in #270
- Ibus refactor2 by @tokuhirom in #272
- SKK-JISYO.L をデフォルトで読み込むようにする Close #258 by @tokuhirom in #273
- ibus-akaza をリファクタリングしたことによりエンバグした部分を修正 by @tokuhirom in #274
- [非互換] コストの計算を利用時にするように変更。 by @tokuhirom in #275
- 平仮名を優先する。 #252 by @tokuhirom in #276
- add TUT romkan map #239 by @tokuhirom in #277
- ライブコンバージョンモードでも、lookup table by @tokuhirom in #278
- refactor keymap by @tokuhirom in #279
- add kana-preferred mode for akaza-data tokenizer. by @tokuhirom in #280
- make a directory before install files by @tkng in #283
- refactor akaza-conf by @tokuhirom in #284
- User dict by @tokuhirom in #286
- change keyboard layout to "default" by @tkng in #287
- Revert "平仮名を優先する。 #252" by @tokuhirom in #290
- add executable bit for ibus-akaza installation by @tokuhirom in #292
- fix the current directory retrieving feature by @tokuhirom in #295
- Fix rustfmt formatting issues by @tokuhirom in #301
- Separate lint and test jobs in CI workflow by @tokuhirom in #304
- Pin Rust version to 1.67.0 by @tokuhirom in #306
- Fix clippy warnings for Rust 1.67.0 by @tokuhirom in #307
- Upgrade Rust to 1.68.0 by @tokuhirom in #309
- Upgrade Rust to 1.69.0 by @tokuhirom in #310
- Upgrade Rust to 1.70.0 by @tokuhirom in #311
- Upgrade Rust to 1.71.0 by @tokuhirom in #312
- Upgrade Rust to 1.72.0 by @tokuhirom in #313
- Upgrade Rust to 1.73.0 by @tokuhirom in #314
- Upgrade Rust from 1.73.0 to 1.74.0 by @tokuhirom in #315
- Upgrade Rust from 1.74.0 to 1.75.0 by @tokuhirom in #316
- Upgrade Rust from 1.75.0 to 1.77.0 by @tokuhirom in #317
- Add cargo-deny job to CI by @tokuhirom in #318
- Bump rustix from 0.36.9 to 0.36.16 by @dependabot[bot] in #296
- Bump unsafe-libyaml from 0.2.7 to 0.2.10 by @dependabot[bot] in #297
- Upgrade Rust from 1.77.0 to 1.81.0 by @tokuhirom in #320
- Upgrade Rust from 1.81.0 to 1.88.0 by @tokuhirom in #321
- Fix cargo-deny by installing manually by @tokuhirom in #319
- Upgrade Rust from 1.88.0 to 1.92.0 by @tokuhirom in #322
- Update dependencies to latest compatible versions by @tokuhirom in #300
- Fix cargo-deny: Add Apache-2.0 WITH LLVM-exception to allowed licenses by @tokuhirom in #323
- Add Renovate configuration by @tokuhirom in #325
- Add Unicode-3.0 to allowed licenses by @tokuhirom in #324
- Update kelp from 0.4.0 to 0.6.0 to remove atty dependency by @tokuhirom in #327
- Update vibrato from 0.3.3 to 0.4.0 by @tokuhirom in #328
- Add workspace resolver = "2" setting by @tokuhirom in #298
- Update vibrato from 0.4.0 to 0.5.0 by @tokuhirom in #329
- Update vibrato to 0.5.2 by @tokuhirom in #330
- Remove unnecessary extern crate declarations by @tokuhirom in #299
- Update GTK4 dependencies to 0.6.6 by @tokuhirom in #331
- Update GTK4 dependencies to 0.7.3 by @tokuhirom in #332
- Update GTK4 dependencies to 0.8.2 by @tokuhirom in #333
- Update GTK4 dependencies to 0.9.7 by @tokuhirom in #334
- Update GTK4 dependencies from 0.9.7 to 0.10.3 by @tokuhirom in #335
- Update dependencies to latest versions by @tokuhirom in #336
- Add tagpr and update README by @tokuhirom in #337
- Update README with comprehensive improvements by @tokuhirom in #338
- Fix tagpr workflow to use GitHub App token by @tokuhirom in #339
- Update renovate.json by @tokuhirom in #340
- Make Renovate more aggressive with frequent updates by @tokuhirom in #341
v0.1.7 - 2026-01-27
- Update GTK4 dependencies from 0.5.5 to 0.10.3
- Update common dependencies to latest versions
- Fix env_logger 0.11 API compatibility