At a Glance
The project in brief
BASED ON
Zadd Translate’s move to human-aided machine translation
PERIOD
2019–2022
CATEGORY
Translation & Interpretation
TOOL
MemoQ
TRAINING TIME
Two years before full launch
SET-UP
A fully active memory, one per client
The Situation
Machine translation was coming either way
Machine translation was changing the market, and Zadd had to adapt. The risk was switching too early, and handing demanding clients output that wasn’t good enough.
the method
Train first, launch later, and keep each client separate
- Train for two years. We trained the model in MemoQ for two years, while translators kept working as usual, before it went fully live.
- Launch with a fully active memory. When it did go live, the memory was already rich enough to help from day one.
- One memory per client. Every client got their own memory, so each one’s terminology and voice stayed consistent and never bled into another’s work.
- Keep humans in charge. The machine drafted and suggested. Translators decided.
the lessons
What any translation team can take from it
- Machine translation is only as good as the memory behind it. Budget the time to build it.
- Separate memories per client protect consistency, and your clients’ trust.
- Human-aided beats machine-only when the material is technical, legal or high-stakes.
The outcome is in the success story, Moving Zadd Translate to Human-Aided Machine Translation.
Work with me
Planning coverage in Jordan?
I bring the access, the interpretation and the local knowledge, and I work to your deadline.