fix: IPA生成でエ段の母音をɛからeに変更しGoogle TTSの発音を修正#1418
Merged
TinyKitten merged 4 commits intodevfrom Mar 6, 2026
Merged
Conversation
Contributor
📝 WalkthroughWalkthrough片仮名→IPA変換の内部ロジックを修正し、複数母音マッピングで Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Suggested labels
Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches
🧪 Generate unit tests (beta)
Comment |
…pronunciation The open-mid front vowel ɛ was causing Google TTS to misinterpret Japanese エ-row sounds. For example, セイブ (Seibu/西武) generated IPA "sɛibɯ" which TTS read as "さいぶ" instead of "せいぶ". Changes: - Replace all ɛ mappings with e (mid front vowel) for エ-row kana - Add ei → eː vowel contraction rule (parallel to existing oɯ → oː) so エイ sequences produce long [eː] as in natural Japanese speech https://claude.ai/code/session_0199yco9qEy3UF4UoBegNEzN
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@stationapi/src/domain/ipa.rs`:
- Around line 382-388: Unit tests (test_seibu, test_keisei, test_meitetsu)
verify the ei → eː IPA mapping in ipa.rs, but you must validate the real Google
TTS output before merging: run the IPA conversion for those three station names
using the conversion function in ipa.rs, synthesize audio via Google TTS with
the same locale/voice settings used in production, save the generated audio, and
compare the perceived pronunciations against the expected IPA (note any
mismatches such as whether Google renders "ei" as a long e or as separate
vowels); then attach the audio samples and a short result summary to the PR and,
if Google differs, update the ipa.rs mapping or add a normalization/SSML
workaround so runtime TTS matches the intended pronunciation.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: dccb0518-6a80-42bb-a2b7-a4a56380c35c
📒 Files selected for processing (1)
stationapi/src/domain/ipa.rs
Update the new name_ipa tests added in dev to match the ɛ→e vowel change. https://claude.ai/code/session_0199yco9qEy3UF4UoBegNEzN
TinyKitten
force-pushed
the
claude/fix-seibu-ipa-pronunciation-kpjNL
branch
from
0c09cfe to
2d1a027
Compare
TinyKitten
commented
Mar 6, 2026
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@stationapi/src/domain/ipa.rs`:
- Around line 382-388: The ei→eː branch can produce duplicate long-tone marks
(e.g. input "...eiー" -> "eːː"); modify the block in ipa.rs that handles the `if
chars[i] == 'e' && chars[i + 1] == 'i'` case so that after you intend to append
'ː' you first check and avoid duplicating it — e.g. skip pushing 'ː' if the last
char in `result` is already 'ː' or if the next input char (chars[i+2] if
present) is a chōon mark like 'ː' or 'ー'; update the `result.push('e');
result.push('ː'); i += 2;` logic in that branch accordingly.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: ca9f833b-84c1-4815-81d7-af44526c0984
📒 Files selected for processing (2)
stationapi/src/domain/ipa.rsstationapi/src/use_case/dto/line.rs
When input like "eiː" is processed by apply_vowel_length, the ei→eː rule already appends ː, then the trailing ː would be pushed again. Skip a following long-vowel mark after the contraction to avoid "eːː". https://claude.ai/code/session_0199yco9qEy3UF4UoBegNEzN
Contributor
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
stationapi/src/domain/ipa.rs (1)
368-392:⚠️ Potential issue | 🟡 Minor
ei以外の長母音収縮でːが重複する余地があります。
ei分岐では後続長音を吸収していますが、oɯ -> oː/oo -> oːでは同じ吸収がなく、入力次第でoːːが残る可能性があります。収縮分岐で処理を統一したほうが安全です。🔧 提案差分
while i < len { if i + 1 < len && chars[i] == 'o' && chars[i + 1] == 'ɯ' { // oɯ → oː (おう/こう pattern) result.push('o'); result.push('ː'); i += 2; + if i < len && chars[i] == 'ː' { + i += 1; + } continue; } if i + 1 < len && chars[i] == 'o' && chars[i + 1] == 'o' { // oo → oː (おお pattern) result.push('o'); result.push('ː'); i += 2; + if i < len && chars[i] == 'ː' { + i += 1; + } continue; } if i + 1 < len && chars[i] == 'e' && chars[i + 1] == 'i' { // ei → eː (えい/けい pattern — 京成 keisei → keːseː) result.push('e'); result.push('ː'); i += 2; // Skip a following long-vowel mark to avoid duplicate 'ː' - if i < len && (chars[i] == 'ː' || chars[i] == 'ー') { + if i < len && chars[i] == 'ː' { i += 1; } continue; }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@stationapi/src/domain/ipa.rs` around lines 368 - 392, The oɯ and oo long-vowel branches can leave a duplicate 'ː' because they don't skip a following long-vowel mark like the ei branch does; update the if blocks that check chars[i]=='o' && chars[i+1]=='ɯ' and chars[i]=='o' && chars[i+1]=='o' so after pushing 'o' and 'ː' and advancing i by 2 you also check if i < len and (chars[i] == 'ː' || chars[i] == 'ー') and, if so, increment i by 1 to consume the existing long-vowel marker (use the same pattern as the ei branch; refer to variables chars, i, len, and result in ipa.rs).
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Outside diff comments:
In `@stationapi/src/domain/ipa.rs`:
- Around line 368-392: The oɯ and oo long-vowel branches can leave a duplicate
'ː' because they don't skip a following long-vowel mark like the ei branch does;
update the if blocks that check chars[i]=='o' && chars[i+1]=='ɯ' and
chars[i]=='o' && chars[i+1]=='o' so after pushing 'o' and 'ː' and advancing i by
2 you also check if i < len and (chars[i] == 'ː' || chars[i] == 'ー') and, if so,
increment i by 1 to consume the existing long-vowel marker (use the same pattern
as the ei branch; refer to variables chars, i, len, and result in ipa.rs).
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 554f4243-3713-4cd1-a187-d6e1c4ed9c8c
📒 Files selected for processing (1)
stationapi/src/domain/ipa.rs
Same issue as the ei→eː branch: if the input already has a trailing long-vowel mark after the digraph, the contraction would produce "oːː". Add the same skip-next-ː guard to both branches. https://claude.ai/code/session_0199yco9qEy3UF4UoBegNEzN
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (1)
stationapi/src/domain/ipa.rs (1)
373-398:'ー'のチェックは到達不能コードの可能性があります。
apply_vowel_lengthが呼ばれる時点で、全ての'ー'は既にapply_phonological_rules内のPhoneme::LongVowel処理により'ː'に変換されています。そのため、chars[i] == 'ー'の条件は常にfalseになります。防御的プログラミングとして残すか、削除するかはお任せします。
♻️ 簡略化案（オプショナル）
// Skip a following long-vowel mark to avoid duplicate 'ː' - if i < len && (chars[i] == 'ː' || chars[i] == 'ー') { + if i < len && chars[i] == 'ː' { i += 1; }同様の変更を3箇所（lines 374, 385, 396）に適用。
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@stationapi/src/domain/ipa.rs` around lines 373 - 398, apply_vowel_length contains redundant checks for 'ー' because apply_phonological_rules already converts Phoneme::LongVowel to 'ː'; remove the unreachable chars[i] == 'ー' conditions in the three places inside apply_vowel_length (the blocks handling "aa", "oo", and "ei") so the skip-after-long-vowel checks only test for 'ː' (or, alternatively, leave a single explanatory comment and keep only 'ː' for defensive clarity); locate these checks by searching for the comments "Skip a following long-vowel mark to avoid duplicate 'ː'" within apply_vowel_length and update the conditions accordingly.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@stationapi/src/domain/ipa.rs`:
- Around line 373-398: apply_vowel_length contains redundant checks for 'ー'
because apply_phonological_rules already converts Phoneme::LongVowel to 'ː';
remove the unreachable chars[i] == 'ー' conditions in the three places inside
apply_vowel_length (the blocks handling "aa", "oo", and "ei") so the
skip-after-long-vowel checks only test for 'ː' (or, alternatively, leave a
single explanatory comment and keep only 'ː' for defensive clarity); locate
these checks by searching for the comments "Skip a following long-vowel mark to
avoid duplicate 'ː'" within apply_vowel_length and update the conditions
accordingly.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 74aec398-184f-4833-be3e-720025b1adf4
📒 Files selected for processing (1)
stationapi/src/domain/ipa.rs
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
ɛを中段前舌母音eに変更ɛiを英語の /aɪ/ のように解釈し、「西武(セイブ)」が「さいぶ」と発音される問題を修正ei → eːの長母音化ルールを追加（既存のoɯ → oːと同様のパターン）Test plan
https://claude.ai/code/session_0199yco9qEy3UF4UoBegNEzN
Summary by CodeRabbit
リリースノート
バグ修正
テスト