English/kouyu_english/lib/core/speech_compare.dart

/// 把设备转写和示范句逐词对照。这不是发音评分：转写只说明“机器听成了什么”，
/// 用来提示学习者哪些词没被听出来，值得再听示范、再说一次。
class SpokenWord {
  const SpokenWord(this.text, {required this.heard});

  /// 示范句中的原词（保留大小写和标点，便于直接显示）。
  final String text;
  final bool heard;
}

class SpokenComparison {
  const SpokenComparison({required this.words, required this.extraWords});

  final List<SpokenWord> words;

  /// 转写里出现、但示范句没有的词。
  final List<String> extraWords;

  int get heardCount => words.where((word) => word.heard).length;
  int get total => words.length;
  List<String> get missedWords => [
    for (final word in words)
      if (!word.heard) word.text,
  ];

  /// 全部听出且没有多余词。
  bool get matches => total > 0 && heardCount == total && extraWords.isEmpty;

  /// 大部分词被听出（≥ 80%）。
  bool get close => total > 0 && heardCount / total >= 0.8;
}

const _digitWords = [
  'zero',
  'one',
  'two',
  'three',
  'four',
  'five',
  'six',
  'seven',
  'eight',
  'nine',
];

const _contractions = {
  "i'm": ['i', 'am'],
  "it's": ['it', 'is'],
  "what's": ['what', 'is'],
  "that's": ['that', 'is'],
  "he's": ['he', 'is'],
  "she's": ['she', 'is'],
  "name's": ['name', 'is'],
  "don't": ['do', 'not'],
  "you're": ['you', 'are'],
};

/// 一个显示词对应的比较单位。例如 "I’m" 展开为 i/am，"138" 展开为
/// one/three/eight，"o’clock" 统一为 oclock。
List<String> _unitsOf(String raw) {
  var word = raw
      .toLowerCase()
      .replaceAll(RegExp(r'[‘’`´]'), "'")
      .replaceAll(RegExp(r"[^a-z0-9']"), '');
  word = word.replaceAll(RegExp(r"^'+|'+$"), '');
  if (word.isEmpty) return const [];
  if (word == "o'clock") return const ['oclock'];
  final expanded = _contractions[word];
  if (expanded != null) return expanded;
  if (RegExp(r'^\d+$').hasMatch(word)) {
    return [for (final digit in word.split('')) _digitWords[int.parse(digit)]];
  }
  return [word.replaceAll("'", '')];
}

List<String> _displayWords(String text) => text
    .replaceAll(RegExp(r'[-–—/]'), ' ')
    .split(RegExp(r'\s+'))
    .where((word) => _unitsOf(word).isNotEmpty)
    .toList();

List<String> _heardUnits(String transcript) => [
  for (final word in _displayWords(transcript)) ..._unitsOf(word),
];

SpokenComparison compareSpoken(String expected, String transcript) {
  final display = _displayWords(expected);
  final expectedUnits = <String>[];
  final owner = <int>[];
  for (var index = 0; index < display.length; index++) {
    for (final unit in _unitsOf(display[index])) {
      expectedUnits.add(unit);
      owner.add(index);
    }
  }
  var heard = _heardUnits(transcript);
  // 拼读时转写常把字母连成一个词（S H E N → Shen），拆回单个字母。
  final spellsLetters =
      expectedUnits.isNotEmpty &&
      expectedUnits.every((unit) => RegExp(r'^[a-z]$').hasMatch(unit));
  if (spellsLetters) {
    heard = [for (final unit in heard) ...unit.split('')];
  }

  // 最长公共子序列：保持词序，重复词也能正确对齐。
  final n = expectedUnits.length;
  final m = heard.length;
  final table = List.generate(n + 1, (_) => List.filled(m + 1, 0));
  for (var i = n - 1; i >= 0; i--) {
    for (var j = m - 1; j >= 0; j--) {
      table[i][j] = expectedUnits[i] == heard[j]
          ? table[i + 1][j + 1] + 1
          : (table[i + 1][j] >= table[i][j + 1]
                ? table[i + 1][j]
                : table[i][j + 1]);
    }
  }
  final unitHeard = List.filled(n, false);
  final heardUsed = List.filled(m, false);
  var i = 0;
  var j = 0;
  while (i < n && j < m) {
    if (expectedUnits[i] == heard[j]) {
      unitHeard[i] = true;
      heardUsed[j] = true;
      i++;
      j++;
    } else if (table[i + 1][j] >= table[i][j + 1]) {
      i++;
    } else {
      j++;
    }
  }

  final wordHeard = List.filled(display.length, true);
  for (var unit = 0; unit < n; unit++) {
    if (!unitHeard[unit]) wordHeard[owner[unit]] = false;
  }
  return SpokenComparison(
    words: [
      for (var index = 0; index < display.length; index++)
        SpokenWord(display[index], heard: wordHeard[index]),
    ],
    extraWords: [
      for (var index = 0; index < m; index++)
        if (!heardUsed[index]) heard[index],
    ],
  );
}