英語スピーキングシャドーイングシステム設計：90日ルーティンと測定指標

# 90日シャドーイングプログラム設定
program:
  name: '90-Day Shadowing System'
  daily_minutes: 15
  phases:
    - name: 'Foundation'
      weeks: 1-2
      stage: 'Mumbling'
      material: 'VOA Learning English'
      daily_routine:
        - listen_without_speaking: 3min
        - mumbling_practice: 10min
        - vocabulary_review: 2min
      target: 'リズムについていく割合 40%以上'

    - name: 'Sync'
      weeks: 3-4
      stage: 'Synchronized'
      material: 'TED-Ed'
      daily_routine:
        - full_speed_listening: 2min
        - synchronized_shadowing: 10min
        - script_check: 3min
      target: '同期率 60%以上'

    - name: 'Prosody'
      weeks: 5-8
      stage: 'Prosody Shadowing'
      material: 'TED Talks'
      daily_routine:
        - shadowing_with_recording: 10min
        - compare_with_original: 3min
        - note_problem_sounds: 2min
      target: 'イントネーション類似度 自己評価 7/10以上'

    - name: 'Content'
      weeks: 9-12
      stage: 'Content Shadowing'
      material: 'NPR / BBC'
      daily_routine:
        - blind_shadowing: 8min
        - retelling_in_own_words: 5min
        - new_vocabulary_log: 2min
      target: '初回シャドーイング成功率 70%以上'

[0:00-0:30]  今日の素材を再生 - 全体1回聴取（要点把握）
[0:30-2:00]  1回目シャドーイング - 録音開始
[2:00-4:00]  録音再生 + 原音と比較 - 差異3つをメモ
[4:00-8:00]  問題箇所を集中反復（最低5回）
[8:00-12:00] 2回目シャドーイング - 全体を通して録音
[12:00-14:00] 2回目録音 vs 原音比較 - 改善点を確認
[14:00-15:00] 今日学んだ表現3つをメモ（単語帳アプリに追加）

import difflib
from google.cloud import speech_v1 as speech

def measure_accuracy(audio_path: str, expected_text: str) -> dict:
    """シャドーイング録音ファイルの発音精度を測定する。"""
    client = speech.SpeechClient()

    with open(audio_path, "rb") as f:
        audio = speech.RecognitionAudio(content=f.read())

    config = speech.RecognitionConfig(
        encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz=16000,
        language_code="en-US",
        enable_word_time_offsets=True,
    )

    response = client.recognize(config=config, audio=audio)

    recognized = " ".join(
        result.alternatives[0].transcript
        for result in response.results
    ).lower().split()

    expected = expected_text.lower().split()

    # 単語レベルの一致率を計算
    matcher = difflib.SequenceMatcher(None, expected, recognized)
    accuracy = matcher.ratio() * 100

    # 聞き取れなかった単語リストを抽出
    missed_words = []
    for tag, i1, i2, j1, j2 in matcher.get_opcodes():
        if tag in ("replace", "delete"):
            missed_words.extend(expected[i1:i2])

    return {
        "accuracy_pct": round(accuracy, 1),
        "total_expected": len(expected),
        "total_recognized": len(recognized),
        "missed_words": missed_words[:10],  # 上位10個のみ
    }

# 使用例
result = measure_accuracy(
    "recordings/2026-03-04-ted-talk.wav",
    "The greatest glory in living lies not in never falling "
    "but in rising every time we fall"
)
print(f"精度: {result['accuracy_pct']}%")
print(f"聞き取れなかった単語: {result['missed_words']}")

import json
from datetime import date, timedelta

def weekly_log(week_num: int, data: dict) -> dict:
    """週間シャドーイング成果を記録し、前週と比較する。"""
    log = {
        "week": week_num,
        "period": f"{date.today() - timedelta(days=6)} ~ {date.today()}",
        "metrics": {
            "wpm": data.get("wpm", 0),
            "sync_rate_pct": data.get("sync_rate", 0),
            "pronunciation_accuracy_pct": data.get("accuracy", 0),
            "retelling_score": data.get("retelling", 0),
        },
        "practice_days": data.get("days_practiced", 0),
        "total_minutes": data.get("total_minutes", 0),
        "material_used": data.get("materials", []),
        "problem_sounds": data.get("problem_sounds", []),
        "notes": data.get("notes", ""),
    }

    # JSONファイルに累積保存
    log_file = "shadowing_progress.json"
    try:
        with open(log_file, "r") as f:
            all_logs = json.load(f)
    except FileNotFoundError:
        all_logs = []

    all_logs.append(log)

    with open(log_file, "w") as f:
        json.dump(all_logs, f, indent=2, ensure_ascii=False)

    return log

# 使用例
weekly_log(5, {
    "wpm": 105,
    "sync_rate": 65,
    "accuracy": 72,
    "retelling": 5,
    "days_practiced": 6,
    "total_minutes": 90,
    "materials": ["TED: The power of vulnerability"],
    "problem_sounds": ["th sound", "r/l distinction", "word-final consonants"],
    "notes": "リエゾン処理がかなり改善されたが、th発音はまだ難しい",
})