Machine translation post-editing

What post-editing is

The text is run through a machine translation engine and a professional translator then corrects the output. Light post-editing fixes errors of meaning, omissions and anything misleading, and leaves awkward but correct sentences alone; the result is understandable, not polished. Full post-editing brings the text to the standard of human translation in accuracy, terminology and readability, following the practice described in ISO 18587 (we follow that practice; we do not hold ISO certification). The saving comes from the translator editing rather than composing, which for suitable text is faster.

Where it works between Turkish and English

• Internal documentation, meeting notes, intranet content and policies that staff need to understand but that nobody will publish.
• Large product catalogues and the long tail of e-commerce descriptions, after the top sellers have been translated by hand.
• Support tickets, chat logs, reviews and other user-generated content for analysis or triage.
• Legacy help articles and knowledge bases.
• Discovery-stage document review in litigation, to decide which documents deserve full translation.
• First drafts of long technical texts with a stable glossary, followed by full post-editing.

Where it does not work, and we will say so

• Anything to be certified for a UK or Turkish authority. A certified translation carries a statement that we made and checked it; that has to be true.
• Contracts for signature, court submissions, notarial documents.
• Marketing, brand and website copy meant to persuade.
• Safety instructions, clinical records for treatment, dosage information.
• Text with personal or special-category data that cannot be processed by a third-party engine under your data-protection obligations.

Turkish is a hard language for engines: suffix agreement, formality (siz and sen), the reverse sentence order and legal and administrative terminology are where errors concentrate, and these are the errors that look fluent and are hardest to spot. English to Turkish output of general prose is now usable as a draft; Turkish to English output of legal and administrative text is often not, and post-editing it can take longer than translating.

Data handling

We use engines under a data-processing agreement with no retention of your text for training, not free public tools. Files with personal data are pseudonymised where feasible before processing. If your contract with your own client forbids machine processing, tell us and we will translate by hand. Our data protection policy describes the controls.

What it costs

Quoted per project below our human translation rates, with the difference depending on how good the raw output is for your text type: we run a sample first and price on what we find. Light post-editing of suitable content is the cheapest tier we offer; full post-editing of difficult text can approach the cost of translation, in which case we recommend translation instead. Translation memory and your glossary are applied to the engine output before editing to cut the work further. No UK VAT is charged.

Turnaround

Roughly twice the daily throughput of translation for suitable text: a 50,000-word knowledge base in about two weeks with full post-editing, less with light. A sample of 1,000 words is turned round in a day so you can judge the tier before committing.

How to decide

Ask what happens if a sentence is wrong. If the answer is “someone asks a colleague”, post-editing is probably fine. If the answer involves a court, a regulator, a patient or a customer’s trust, it is not. Our revision service is the alternative for text you have already had translated by other means.