Persona C — Translation Tooling

Scenario

You build a CAT tool (computer-assisted translation) or TMS (translation memory system) for e-commerce content. Your customers translate product descriptions, category trees, customer support articles. Generic TM hits ~60% — you need higher.

What mogan-i18n replaces

• Generic translation memory match (~60% e-commerce hit rate)
• Per-customer training of TM with their catalog (months of pipeline work)
• Manual category mapping QA across locales

Working code (Python pipeline)

# Augment your TM with mogan-i18n cross-locale anchor
import requests

def augment_tm_segment(source_text: str, source_locale: str, target_locale: str):
    # 1. Try mogan-i18n canonical lookup first
    r = requests.get(
        "https://api.mogan-i18n.com/v1/lookup",
        params={"keyword": source_text, "locale": source_locale},
        headers={"x-api-key": "your_key"}
    )

    if r.status_code == 200:
        canonical = r.json()["canonical_slug"]
        # Get all locale variants
        variants = requests.get(
            f"https://api.mogan-i18n.com/v1/canonical/{canonical}",
            headers={"x-api-key": "your_key"}
        ).json()

        if target_locale in variants["variants"]:
            return {
                "translation": variants["variants"][target_locale],
                "source": "mogan-i18n",
                "confidence": variants["confidence"],
                "_provenance": variants["_provenance"]
            }

    # 2. Fall back to your existing TM
    return your_tm.lookup(source_text, source_locale, target_locale)

Pipeline integration sketch

Insert mogan-i18n at the segment level before your TM/MT fallback. For e-commerce content, expect 70-80% canonical hit rate (vs 60% generic TM). Misses gracefully fall through your existing pipeline. Each canonical hit ships with confidence + Jaccard back-translation validation score for QA gates.