Domain Customization
We tune MT engines to your industry — legal, medical, financial — so output respects the terminology, register, and conventions your content demands.
TECHNOLOGY SOLUTIONS · AI-POWERED MACHINE TRANSLATION
Machine translation,tuned to your domain.
A managed service built on Arabic.Ai LLM: we deploy, customize, and train MT engines for your domain — legal, medical, financial — with post-editing pipelines that combine machine output and expert human review.
Overview
Raw MT isn't enterprise-ready.
Off-the-shelf machine translation gets you a draft, not a deliverable — especially in regulated, high-stakes domains. Our AI-powered machine translation service is built on Arabic.Ai LLM and delivered as a managed engagement: we deploy and customize MT engines for your specific domain (legal, medical, financial), train them on your terminology and past translations, and wrap them in post-editing pipelines where expert human linguists review and refine the machine output. You get the speed and scale of MT with the accuracy, consistency, and cultural nuance that only domain-tuned engines plus human review can provide.
What's Included
Arabic AI built for business reality.
Engine Training
Engines are trained on your terminology, glossaries, and past translations, so the model learns your voice and improves with every project.
Post-Editing Pipelines
Machine output flows into structured post-editing where expert linguists review, correct, and certify — combining MT speed with human accuracy.
Arabic-First Quality
Built on Arabic.Ai LLM with native MSA and dialect handling, Arabic-typography-aware checks, and cultural-relevance review across 100+ languages.
Managed Deployment
We deploy, host, and maintain the engines for you — via API or in your environment — with monitoring, retraining, and version control handled end-to-end.
Quality Scoring
Confidence scoring, post-editing distance metrics, and human evaluation let you measure quality objectively and decide where review effort is best spent.
Proof
Trusted where it can't go wrong.
E-commerce
Noon: 45% cost reduction with 3× capacity uplift on Arabic localization
Noon, a leading MENA e-commerce platform, needed Arabic localization across a vast and constantly-changing product catalog. Tarjama deployed 100+ linguists and an optimized TMS to triple throughput while cutting cost.
45%cost reduction
3×capacity uplift
100+linguists
How it works
AI machine translation, done responsibly.
Machine translation delivers value only when it is matched to the right content and backed by human oversight. Here is how Tarjama deploys it.
Step 1
Assess your content & goals
We identify your content types, language pairs, and quality targets to decide where MT fits and where human translation is required.
Step 2
Engine selection & customization
We select and adapt machine-translation engines, training on your data and glossaries for domain-accurate output.
Step 3
Integration
MT is connected into your workflow through API or TMS so it runs where your content already lives.
Step 4
Post-editing setup
We define post-editing levels — light or full — per content type, following ISO 18587 for human review of machine output.
Step 5
Monitor & improve
We track quality metrics and retrain engines over time so output keeps getting better.
FAQ
Questions, answered.
Off-the-shelf engines give you a generic draft. We deploy and customize MT for your specific domain, train it on your terminology and past work, and add expert post-editing — so the output is accurate, consistent, and publishable, not just readable.
Any specialized field — legal, medical, financial, technical, government — where terminology and register matter. The engine learns from your glossaries, reference material, and translation memory.
After the engine produces a translation, expert human linguists review and refine it against your standards. You can choose light post-editing for speed or full post-editing for publication-grade quality.
No. It's Arabic-first — built on Arabic.Ai LLM with deep MSA and dialect handling — but supports 100+ languages, with the same domain-tuning and post-editing approach across all of them.
Yes. We run a benchmark on a sample of your real content so you can see domain-tuned MT plus post-editing quality before committing to a full engagement.
Ready to start?
Tune MT to your domain.
Tell us your use case and we'll set up a 30-day evaluation with enough credits to benchmark properly — plus support to make sure the test is a fair fight.