Skip to content

Glossary Structure

Falara glossaries support multi-language terminology management with disambiguation, domain scoping, and forbidden alternative enforcement.

Entry Structure

A glossary entry maps a source term to translations in one or more target languages:

{
  "id": "term_f3a1b2c3",
  "term": "Hypertonie",
  "translations": {
    "en": "hypertension",
    "fr": "hypertension",
    "es": "hipertensión"
  },
  "context": "cardiovascular condition — not software pressure",
  "domain": "medical",
  "forbidden_alternatives": ["high blood pressure", "high BP"],
  "status": "approved"
}

Field Reference

term

The source-language term to match. Case-insensitive matching is applied.

translations

A dictionary of { target_language_code: translated_term }. At least one entry is required.

Entries are applied only when the translation target language matches a key in this dictionary. A single glossary can cover any number of target languages.

context

Optional natural-language description that disambiguates polysemous terms. The context is passed to the Translator agent to ensure the correct sense is used.

Example: "banking context — not gambling" for the term stake.

domain

Subject domain label, e.g. "medical", "legal", "finance". Helps the Translator apply domain-appropriate register.

forbidden_alternatives

A list of terms the Translator is explicitly instructed to avoid. Useful for enforcing house style or brand terminology.

Example: For term "hypertension", forbidden: ["Hypertonie"] ensures the translator uses "Bluthochdruck" in German and does not fall back to the Latin form.

status

Value Meaning
approved Term is validated and actively enforced
pending Draft — included in translation requests but flagged for review
deprecated No longer enforced; kept for historical reference

Size Recommendations

Size Recommendation
500–2,000 entries Optimal for single-domain glossaries
Up to 5,000 entries Supported; performance may slightly decrease
> 2,000 entries Consider splitting by domain into separate glossaries

Multi-Language Example

A single glossary covering EN→DE and EN→FR:

{
  "name": "E-Commerce Terms",
  "source_lang": "en",
  "entries": [
    {
      "term": "shopping cart",
      "translations": { "de": "Warenkorb", "fr": "panier" },
      "domain": "e-commerce"
    },
    {
      "term": "checkout",
      "translations": { "de": "Kasse", "fr": "paiement" },
      "forbidden_alternatives": { "de": ["Auschecken"], "fr": ["sortie"] },
      "domain": "e-commerce"
    }
  ]
}

When used in a DE translation job, only the "de" translations are applied. Same glossary works for FR jobs without duplication.