AI Content Translation Implementation

AI-Powered Content Translation for Websites

Automated website content translation to multiple languages using neural models. Fast, cost-effective alternative to human translation for high-volume sites.

Translation Approaches

Direct API (Google Translate, DeepL) — fastest, suitable for non-critical content. Cost: $15–20 per million characters.

Fine-tuned models (Azure Translator, AWS Translate) — better for domain-specific terminology. 2–3 week setup.

Open-source (Argos, Ollama + Seamless M4T) — self-hosted, no per-request costs, full control.

Implementation via DeepL API

import * as deepl from 'deepl-node';

const translator = new deepl.Translator(process.env.DEEPL_API_KEY);

async function translateContent(text, targetLang = 'EN') {
  const result = await translator.translateText(text, null, targetLang);
  return result.text;
}

// Batch translation
async function translatePage(page, languages = ['DE', 'FR', 'UK']) {
  const translations = {};

  for (const lang of languages) {
    translations[lang] = {
      title: await translateContent(page.title, lang),
      description: await translateContent(page.description, lang),
      body: await translateContent(page.body, lang),
    };
  }

  return translations;
}

// Save to database
async function saveTranslations(pageId, translations) {
  for (const [lang, content] of Object.entries(translations)) {
    await db.insert('page_translations', {
      page_id: pageId,
      language: lang,
      title: content.title,
      description: content.description,
      body: content.body,
      translated_at: new Date(),
    });
  }
}

SEO & Language Detection

// Auto-detect user language from header
function getPreferredLanguage(request) {
  const acceptLanguage = request.headers['accept-language'];
  if (!acceptLanguage) return 'en';

  const languages = acceptLanguage.split(',').map(lang =>
    lang.split(';')[0].trim().split('-')[0]
  );

  return languages[0] || 'en';
}

// Hreflang tags for SEO
function generateHreflangs(path, languages = ['en', 'de', 'fr', 'uk']) {
  return languages.map(lang =>
    `<link rel="alternate" hreflang="${lang}" href="https://example.com/${lang}${path}" />`
  ).join('\n');
}

Quality Control

// Post-translation QA checks
async function validateTranslation(original, translated, lang) {
  const checks = [
    // Length should be within 20% of original
    Math.abs(translated.length - original.length) / original.length < 0.2,
    // No untranslated placeholders (${var})
    !/{{\w+}}/.test(translated),
    // Proper punctuation
    /[.!?;:]$/.test(translated),
  ];

  return checks.every(c => c);
}

// Use translation API with glossary
async function translateWithGlossary(text, lang, glossary) {
  return translator.translateText(text, null, lang, {
    glossary: glossary,  // Custom term mappings
  });
}

Timeline

• Setup translation API — 1 day
• Batch translate existing content — 2–3 days
• Language detection + routing — 1–2 days
• SEO (hreflang, lang attributes) — 1 day
• Quality review + glossary management — 2–3 weeks ongoing