Function assessTranscriptLanguage

lib/transcript-language.ts:59–114

(
  transcript: unknown,
  options: TranscriptLanguageOptions = {}
)

}

export function assessTranscriptLanguage(
  transcript: unknown,
  options: TranscriptLanguageOptions = {}
): TranscriptLanguageAssessment {
  if (!Array.isArray(transcript) || transcript.length === 0) {
    return { isEnglish: false, reason: 'NO_TRANSCRIPT' };
  }

  const reportedLanguages = transcript
    .map(extractSegmentLanguage)
    .filter((lang): lang is string => Boolean(lang));

  if (reportedLanguages.length > 0) {
    const hasEnglish = reportedLanguages.some(
      (lang) => lang === 'en' || lang.startsWith('en-')
    );
    return {
      isEnglish: hasEnglish,
      reason: hasEnglish ? 'REPORTED_ENGLISH' : 'REPORTED_NON_ENGLISH',
    };
  }

  const sampleLimit = options.sampleSegments ?? DEFAULT_SAMPLE_SEGMENTS;
  const sampleText = transcript
    .slice(0, sampleLimit)
    .map(extractSegmentText)
    .join(' ')
    .replace(/\s+/g, ' ')
    .trim();

  if (!sampleText) {
    return { isEnglish: false, reason: 'EMPTY_SAMPLE' };
  }

  const textWithoutSpaces = sampleText.replace(/\s/g, '');
  const nonSpaceLength = textWithoutSpaces.length;
  if (nonSpaceLength === 0) {
    return { isEnglish: false, reason: 'EMPTY_SAMPLE' };
  }

  const englishLetterCount = (textWithoutSpaces.match(/[A-Za-z]/g) ?? []).length;
  const englishRatio = englishLetterCount / nonSpaceLength;
  const cjkCharacterPresent = CJK_REGEX.test(sampleText);
  const minEnglishRatio = options.minEnglishRatio ?? DEFAULT_MIN_ENGLISH_RATIO;
  const strictCjkThreshold = options.strictCjkThreshold ?? DEFAULT_CJK_RATIO_THRESHOLD;

  if (cjkCharacterPresent && englishRatio < strictCjkThreshold) {
    return { isEnglish: false, reason: 'CJK_DOMINANT', englishRatio };
  }

  if (englishRatio < minEnglishRatio) {
    return { isEnglish: false, reason: 'LOW_ENGLISH_RATIO', englishRatio };
  }

  return { isEnglish: true, reason: 'RATIO_THRESHOLD', englishRatio };
}

export function isTranscriptEnglish(