Skip to content

name_tts_segments・name_ipaの長音をːに統一し語頭ら行をlに修正 #1539

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
TinyKitten merged 1 commit into from
Jun 16, 2026
+133 −43
Merged

name_tts_segments・name_ipaの長音をːに統一し語頭ら行をlに修正 #1539

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
154 changes: 123 additions & 31 deletions stationapi/src/domain/ipa.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1305,20 +1305,34 @@ fn phonemes_to_tagged_chars(phonemes: &[Phoneme]) -> Vec<(char, Option<usize>)>
let mut mora_index = 0usize;
let len = phonemes.len();
let mut i = 0;
// 語頭 (名称先頭・区切り直後) かどうか。語頭のら行 `ɾ` を `l` に置換する判定に使う。
let mut at_word_start = true;

while i < len {
match &phonemes[i] {
Phoneme::Regular(ipa) => {
if ipa.trim().is_empty() {
// Separator (word boundary): not a mora.
output.extend(ipa.chars().map(|c| (c, None)));
at_word_start = true;
} else {
mora_index += 1;
output.extend(ipa.chars().map(|c| (c, Some(mora_index))));
// 語頭のら行 ɾ は Azure ja-JP の `<phoneme alphabet="ipa">` で
// 弱い弾き音になり「ら行」に聞こえにくいため、側面接近音 l に
// 置換する (例: ロッポンギ ɾoppongi→loppongi)。語中の ɾ は弾き音
// のままで自然に鳴るので置換しない。ɾʲ (リョ等) は基底の
// ɾ のみ置換し拗音記号 ʲ は残す。
let lateralize = at_word_start && ipa.starts_with('ɾ');
for (idx, c) in ipa.chars().enumerate() {
let c = if lateralize && idx == 0 { 'l' } else { c };
output.push((c, Some(mora_index)));
}
at_word_start = false;
}
}
Phoneme::MoraicNasal => {
mora_index += 1;
at_word_start = false;
let next_regular = find_next_regular(&phonemes[i + 1..]);
let nasal = match next_regular {
Some(next_ipa) => nasal_for_following(next_ipa),
Expand All @@ -1335,6 +1349,7 @@ fn phonemes_to_tagged_chars(phonemes: &[Phoneme]) -> Vec<(char, Option<usize>)>
}
Phoneme::Geminate => {
mora_index += 1;
at_word_start = false;
if let Some(next_ipa) = find_next_regular(&phonemes[i + 1..]) {
if let Some(c) = geminate_onset_char(next_ipa) {
output.push((c, Some(mora_index)));
Expand All @@ -1343,6 +1358,7 @@ fn phonemes_to_tagged_chars(phonemes: &[Phoneme]) -> Vec<(char, Option<usize>)>
}
Phoneme::LongVowel => {
mora_index += 1;
at_word_start = false;
// Lengthen the preceding vowel unless it is already long.
let already_long = output.last().map(|(c, _)| *c == 'ː').unwrap_or(false);
if !already_long {
Expand Down Expand Up @@ -1378,22 +1394,53 @@ fn starts_with_vowel_or_semivowel(ipa: &str) -> bool {
ipa.starts_with(is_ipa_vowel) || ipa.starts_with('j') || ipa.starts_with('w')
}

/// Whether a `prev`→`cur` vowel sequence forms a Japanese long vowel (長音) per
/// kana orthography:
/// ああ→aː いい→iː うう→ɯː ええ/えい→eː おお/おう→oː
/// `えい` (e→i) and `おう` (o→ɯ) are the orthographic long vowels written with a
/// trailing different kana, so they are included alongside the doubled vowels.
/// Other vowel sequences (ウエ, アイ, アオ, オイ …) are genuine vowel hiatus and
/// are NOT collapsed.
fn forms_long_vowel(prev: char, cur: char) -> bool {
matches!(
(prev, cur),
('a', 'a') | ('i', 'i') | ('ɯ', 'ɯ') | ('e', 'e') | ('e', 'i') | ('o', 'o') | ('o', 'ɯ')
)
}

/// Insert IPA syllable boundary markers (`.`) between consecutive vowels,
/// preserving the per-character mora tags. The inserted `.` is attributed to
/// the preceding character's mora so it never becomes the first character of
/// the following mora (which matters when the accent marker is placed there).
///
/// 連続母音が日本語の長音 (おう/おお/えい等, [`forms_long_vowel`]) を成す場合は
/// 音節境界 `.` ではなく長音記号 `ː` に置き換え、2 つ目の母音を伸ばす。Azure の
/// ja-JP `<phoneme alphabet="ipa">` では `to.ɯkʲo.ɯ` のような母音分割よりも
/// `toːkʲoː` の長音表記のほうが「とーきょー」と自然に読まれる。`ː` は
/// 2 つ目の母音のモーラタグを引き継ぐので、ピッチアクセント層のモーラ数整合は
/// 保たれる (長音 `ー` の扱いと同じ)。
fn insert_syllable_breaks_tagged(input: &[(char, Option<usize>)]) -> Vec<(char, Option<usize>)> {
let mut result: Vec<(char, Option<usize>)> = Vec::with_capacity(input.len());
let mut prev_is_vowel = false;
// 直前の母音の「基底」文字。長音化しても基底母音を保持し、続く母音と再判定する。
let mut prev_vowel: Option<char> = None;

for &(c, tag) in input {
let is_vowel = is_ipa_vowel(c);
if is_vowel && prev_is_vowel {
let prev_tag = result.last().and_then(|(_, t)| *t);
result.push(('.', prev_tag));
if is_ipa_vowel(c) {
if let Some(prev) = prev_vowel {
if forms_long_vowel(prev, c) {
// 2 つ目の母音は ː に置換して伸ばす。基底母音 (prev) は維持。
result.push(('ː', tag));
continue;
}
let prev_tag = result.last().and_then(|(_, t)| *t);
result.push(('.', prev_tag));
}
result.push((c, tag));
prev_vowel = Some(c);
} else {
result.push((c, tag));
prev_vowel = None;
}
result.push((c, tag));
prev_is_vowel = is_vowel;
}

result
Expand Down Expand Up @@ -1459,17 +1506,19 @@ mod tests {
fn test_seibu_shinjuku_line_nasals_are_ascii_n() {
// 西武新宿線: 「新」(ン→ɲ) と「線」(語末ン→ɴ) の両方を n に統一し、
// Azure で双方の「ん」が鳴るようにする (#1536)。
assert_eq!(ipa("セイブシンジュクセン"), "se.ibɯɕindʑɯkɯsen");
assert_eq!(ipa("セイブシンジュクセン"), "seːbɯɕindʑɯkɯsen");
}

#[test]
fn test_osaka() {
assert_eq!(ipa("オオサカ"), "o.osaka");
// オオ → 長音 oː (Azure 向け長音統一)
assert_eq!(ipa("オオサカ"), "oːsaka");
}

#[test]
fn test_kyoto() {
assert_eq!(ipa("キョウト"), "kʲo.ɯto");
// キョウ → 長音 kʲoː (おう→oː)
assert_eq!(ipa("キョウト"), "kʲoːto");
}

#[test]
Expand All @@ -1494,7 +1543,8 @@ mod tests {

#[test]
fn test_ryogoku() {
assert_eq!(ipa("リョウゴク"), "ɾʲo.ɯgokɯ");
// 語頭ら行 ɾʲ→lʲ かつ リョウ→長音 lʲoː
assert_eq!(ipa("リョウゴク"), "lʲoːgokɯ");
}

#[test]
Expand All @@ -1505,7 +1555,8 @@ mod tests {

#[test]
fn test_keisei() {
assert_eq!(ipa("ケイセイ"), "ke.ise.i");
// ケイ・セイ → 長音 keː seː (えい→eː)
assert_eq!(ipa("ケイセイ"), "keːseː");
}

#[test]
Expand All @@ -1515,13 +1566,14 @@ mod tests {

#[test]
fn test_meitetsu() {
// ツ is consistently t͡sɯ (affricate with tie bar)
assert_eq!(ipa("メイテツ"), "me.itet͡sɯ");
// ツ is consistently t͡sɯ (affricate with tie bar)。メイ→長音 meː (えい→eː)
assert_eq!(ipa("メイテツ"), "meːtet͡sɯ");
}

#[test]
fn test_seibu() {
assert_eq!(ipa("セイブ"), "se.ibɯ");
// セイ → 長音 seː (えい→eː)
assert_eq!(ipa("セイブ"), "seːbɯ");
}

#[test]
Expand Down Expand Up @@ -1586,29 +1638,52 @@ mod tests {

#[test]
fn test_itchome() {
assert_eq!(ipa("イッチョウメ"), "itt͡ɕo.ɯme");
// チョウ → 長音 t͡ɕoː (おう→oː)
assert_eq!(ipa("イッチョウメ"), "itt͡ɕoːme");
}

#[test]
fn test_sanchome() {
assert_eq!(ipa("サンチョウメ"), "sant͡ɕo.ɯme");
// チョウ → 長音 t͡ɕoː (おう→oː)
assert_eq!(ipa("サンチョウメ"), "sant͡ɕoːme");
}

#[test]
fn test_koen() {
assert_eq!(ipa("コウエン"), "ko.ɯ.en");
// コウ → 長音 koː、エ は別母音なので音節境界を残す koː.en
assert_eq!(ipa("コウエン"), "koː.en");
}

#[test]
fn test_tokyo() {
assert_eq!(ipa("トウキョウ"), "to.ɯkʲo.ɯ");
// トウ・キョウ → 長音 toː kʲoː (おう→oː)
assert_eq!(ipa("トウキョウ"), "toːkʲoː");
}

#[test]
fn test_nagoya() {
assert_eq!(ipa("ナゴヤ"), "nagoja");
}

#[test]
fn test_roppongi_word_initial_r_is_lateralized() {
// 語頭のら行 ɾ は Azure ja-JP で弾き音が弱く不自然なため l に置換。
// 語中の ɾ (ない) はそのまま。
assert_eq!(ipa("ロッポンギ"), "loppongi");
}

#[test]
fn test_word_initial_r_after_separator_is_lateralized() {
// 区切り直後 (= 後続語の語頭) のら行も l に置換。
assert_eq!(ipa("シン・リンカイ"), "ɕin linka.i");
}

#[test]
fn test_medial_r_stays_flap() {
// 語中のら行は弾き音 ɾ のまま (toɾide / t͡sɯɾɯmi)。
assert_eq!(ipa("トリデ"), "toɾide");
}

#[test]
fn test_sapporo() {
// ッポ → ppo
Expand Down Expand Up @@ -1639,11 +1714,28 @@ mod tests {
assert_eq!(katakana_to_ipa("シブヤX"), None);
}

#[test]
fn test_macron_romaji_converges_with_katakana_long_vowel() {
// マクロン付きローマ字 Ōsaka は ō→オウ と再構成されるが、長音統一により
// オオサカ (カタカナ) と同じ oːsaka に収束する。以前は o.ɯsaka と
// 誤った母音分割になっていた。
assert_eq!(
station_name_to_ipa("オオサカ", Some("Ōsaka")),
Some("oːsaka".to_string())
);
// トウキョウ (おう長音) も toːkʲoː。
assert_eq!(
station_name_to_ipa("トウキョウ", Some("Tōkyō")),
Some("toːkʲoː".to_string())
);
}

#[test]
fn test_station_name_ipa_uses_official_english_wording() {
assert_eq!(
station_name_to_ipa("カサイリンカイコウエン", Some("Kasai-Rinkai Park")),
Some("kasa.i ɾinka.i pɑɹk".to_string())
// 語頭 (空白直後) のら行 Rinkai は ɾ→l
Some("kasa.i linka.i pɑɹk".to_string())
);
}

Expand Down Expand Up @@ -1672,7 +1764,7 @@ mod tests {
);
assert_eq!(
accented_katakana_ipa("京都", "キョウト"),
Some("ˈkʲo.ɯto".to_string())
Some("ˈkʲoːto".to_string())
);
}

Expand Down Expand Up @@ -1701,7 +1793,7 @@ mod tests {
// 「ねりまかすが↓ちょう」= 6 モーラ目 (ガ) の直前に下げ核。
assert_eq!(
accented_katakana_ipa("練馬春日町", "ネリマカスガチョウ"),
Some("neɾimakasɯˈgat͡ɕo.ɯ".to_string())
Some("neɾimakasɯˈgat͡ɕoː".to_string())
);
}

Expand All @@ -1717,7 +1809,7 @@ mod tests {
fn test_station_name_ipa_supports_mixed_english_facility_words() {
assert_eq!(
station_name_to_ipa("トウキョウビッグサイト", Some("Tōkyō Big Sight")),
Some("to.ɯkʲo.ɯ bɪg saɪt".to_string())
Some("toːkʲoː bɪg saɪt".to_string())
);
}

Expand Down Expand Up @@ -1773,20 +1865,20 @@ mod tests {
fn test_station_name_ipa_supports_spaced_romanized_names_from_csv() {
assert_eq!(
station_name_to_ipa("メイテツイチノミヤ", Some("Meitetsu Ichinomiya")),
Some("me.itet͡sɯ it͡ɕinomija".to_string())
Some("meːtet͡sɯ it͡ɕinomija".to_string())
);
}

#[test]
fn test_station_name_ipa_supports_meitetsu_prefixed_station_names_from_csv() {
let cases = [
("メイテツナゴヤ", "Meitetsu Nagoya", "me.itet͡sɯ nagoja"),
("メイテツナゴヤ", "Meitetsu Nagoya", "meːtet͡sɯ nagoja"),
(
"メイテツイチノミヤ",
"Meitetsu Ichinomiya",
"me.itet͡sɯ it͡ɕinomija",
"meːtet͡sɯ it͡ɕinomija",
),
("メイテツギフ", "Meitetsu Gifu", "me.itet͡sɯ giɸɯ"),
("メイテツギフ", "Meitetsu Gifu", "meːtet͡sɯ giɸɯ"),
];

for (katakana, roman, expected) in cases {
Expand All @@ -1801,7 +1893,7 @@ mod tests {
#[test]
fn test_nakaguro_treated_as_word_separator() {
// 中黒「・」は語の区切りとして空白に変換し、全体が None にならないようにする
assert_eq!(ipa("チュウオウ・ソウブ"), "t͡ɕɯ.ɯ.o.ɯ so.ɯbɯ");
assert_eq!(ipa("チュウオウ・ソウブ"), "t͡ɕɯː.oː soːbɯ");
}

#[test]
Expand All @@ -1815,7 +1907,7 @@ mod tests {
// Full-width space between words should be preserved
assert_eq!(
ipa("ドッキョウダイガクマエ　ソウカマツバラ"),
"dokkʲo.ɯda.igakɯma.e so.ɯkamat͡sɯbaɾa"
"dokkʲoːda.igakɯma.e soːkamat͡sɯbaɾa"
);
}

Expand All @@ -1824,7 +1916,7 @@ mod tests {
// Half-width (ASCII) space between words should also be accepted
assert_eq!(
ipa("ドッキョウダイガクマエ ソウカマツバラ"),
"dokkʲo.ɯda.igakɯma.e so.ɯkamat͡sɯbaɾa"
"dokkʲoːda.igakɯma.e soːkamat͡sɯbaɾa"
);
}
}
10 changes: 5 additions & 5 deletions stationapi/src/use_case/dto/line.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -391,7 +391,7 @@ mod tests {
line.line_name_k = "セイブイケブクロセン".to_string();
let grpc_line: GrpcLine = line.into();

assert_eq!(grpc_line.name_ipa, Some("se.ibɯ.ikebɯkɯɾosen".to_string()));
assert_eq!(grpc_line.name_ipa, Some("seːbɯ.ikebɯkɯɾosen".to_string()));
}

#[test]
Expand All @@ -403,7 +403,7 @@ mod tests {
line.line_name_k = "トウカイドウホンセン".to_string();
let grpc_line: GrpcLine = line.into();

assert_eq!(grpc_line.name_ipa, Some("to.ɯka.ido.ɯˈhonsen".to_string()));
assert_eq!(grpc_line.name_ipa, Some("toːka.idoːˈhonsen".to_string()));
}

#[test]
Expand All @@ -413,7 +413,7 @@ mod tests {
line.line_name_k = "トウホクシンカンセン".to_string();
let grpc_line: GrpcLine = line.into();

assert_eq!(grpc_line.name_ipa, Some("to.ɯhokɯɕinˈkansen".to_string()));
assert_eq!(grpc_line.name_ipa, Some("toːhokɯɕinˈkansen".to_string()));
}

#[test]
Expand All @@ -425,7 +425,7 @@ mod tests {
let grpc_line: GrpcLine = line.into();

// name_ipa は日本語読み (下げ核付き)、name_roman_ipa はローマ字 (英語) 読み。
assert_eq!(grpc_line.name_ipa, Some("ke.ise.ˈihonsen".to_string()));
assert_eq!(grpc_line.name_ipa, Some("keːseˈːhonsen".to_string()));
assert_eq!(
grpc_line.name_roman_ipa,
Some("keːseː meɪn laɪn".to_string())
Expand All @@ -441,7 +441,7 @@ mod tests {
line.line_name_r = None;
let grpc_line: GrpcLine = line.into();

assert_eq!(grpc_line.name_ipa, Some("t͡ɕɯ.ɯ.ˈo.ɯ so.ɯbɯsen".to_string()));
assert_eq!(grpc_line.name_ipa, Some("t͡ɕɯː.ˈoː soːbɯsen".to_string()));
assert!(!grpc_line.name_tts_segments.is_empty());
}

Expand Down
Loading
Loading