Service · Post-editing / MTPE

Machine translation post-editing for Indonesian.

Your engine drafts, I bring it to the level you specify: light or full post-editing to ISO 18587, inside your CAT tool, with your glossary applied from segment one.

How a file moves

From engine draft to delivered file, in five steps.

  1. 01

    Engine draft

    DeepL, Google NMT, ModernMT or your custom engine produces the raw output.

  2. 02

    Pre-edit analysis

    Error patterns, terminology gaps and TM leverage checked before editing starts.

  3. 03

    Post-editing

    Accuracy, fluency, register and terminology fixed to the agreed level, LPE or FPE.

  4. 04

    QA review

    Consistency check, client glossary verified, full read-through.

  5. 05

    Delivery

    XLIFF, SDLXLIFF, MQXLIFF or DOCX, with the TM updated.

  • ISO 18587
  • EN → ID and ID → EN
  • TM leverage discounts honoured
  • Trados · memoQ · Phrase

What is MTPE?

Machine translation post-editing services — what agencies need to know

Machine translation post-editing — MTPE, sometimes written PEMT — is the professional process of reviewing and correcting the output of a machine translation engine to bring it up to a defined quality level. As neural machine translation (NMT) engines have become standard in agency workflows, MTPE has moved from a niche service to a core production method for high-volume Indonesian content.

The fundamental distinction between post-editing and human translation from scratch is cost and speed, not quality ceiling. Full post-editing (FPE) of good NMT output, by an experienced post-editor with domain depth, reaches the same standard as human translation — because at that point the process is human translation, with the MT draft serving as a starting point rather than a blank page. Light post-editing (LPE) targets a lower quality threshold intentionally: understandable and accurate, but not polished.

For Indonesian, the quality of MT output varies significantly by domain, formality register, and text type. Legal, financial, and technical Indonesian routinely exposes the structural and terminological limits of general-purpose engines. A post-editor without deep domain knowledge and bilingual proficiency in Indonesian cannot reliably catch or correct these errors. The value of MTPE lies entirely in the expertise of the person doing the editing.

  • DeepL
  • Google NMT
  • ModernMT
  • Microsoft Translator
  • Amazon Translate
  • SYSTRAN
  • Omniscien
  • Trados GroupShare MT
  • Custom engines

Why a legal-certified specialist for MTPE?

Indonesian MT output routinely contains errors that only surface as wrong to a native professional with domain depth — incorrect register, calqued legal terms, false equivalents in financial or technical vocabulary. A general post-editor catches fluency errors; a domain specialist catches accuracy errors that look fluent. The difference matters for content that will be acted upon.

Post-editing levels

Light post-editing vs. full post-editing

ISO 18587 defines two levels of post-editing. Selecting the right level is a project management decision — it depends directly on how the translated content will be used. Both levels are available for Indonesian–English and English–Indonesian.

ISO 18587 — Level 1

Light post-editing

LPE · Lower per-word rate

A functionally accurate, comprehensible output. The goal is intelligibility and factual correctness — not stylistic refinement. Errors that change meaning are corrected; awkward phrasing that does not affect understanding is left as is.

  • Correct any translation errors that change meaning
  • Remove obvious mistranslations and omissions
  • Ensure key terminology is correct
  • Fix critical grammar errors that impede understanding
  • Do not restructure sentences for style or flow
  • Do not adapt to a target-language style guide

Suitable for: Internal communications, rapid-turnaround informational content, reference documentation, draft review, gist understanding of large volumes.

ISO 18587 — Level 2

Full post-editing

FPE · Human-translation standard

Accurate, fluent, terminologically consistent, and appropriate in register. The end reader should not be able to distinguish the output from a translation produced from scratch. The MT draft is a productivity tool, not a quality ceiling.

  • Correct all translation errors and omissions
  • Ensure full linguistic and stylistic fluency in Indonesian or English
  • Apply client glossary and style guide throughout
  • Correct register, formality, and cultural adaptation
  • Restructure sentences where the MT output is technically correct but unnatural
  • Format output consistently with the source document

Suitable for: Published marketing content, client-facing documentation, legal and compliance materials, financial reports, product descriptions, and any content submitted to official bodies.

Which level do you need?

Content that will be published, submitted to a third party, or read by an Indonesian audience representing your brand or client → full PE. Content for internal reference, rapid gist, or draft review → light PE. When in doubt, describe the content type and intended use, and the level will be recommended before any quote is issued.

18587 — The ISO standard every post-editing job is scoped against.

Level agreed at intake, checked at QA, stated on delivery.

How it works

MTPE workflow, step by step

A structured post-editing process keeps quality consistent — whether a single 1,000-word batch or an ongoing high-volume agency workflow.

  1. 01

    Project intake and briefing

    Receive the MT output, source files, client glossary or termbase, style guide, and any domain-specific reference materials. Confirm the MT engine used, post-editing level (LPE or FPE), target quality standard, CAT tool, and delivery format. For ongoing agency projects, a standing brief covering all these parameters is agreed once and applied across batches.

  2. 02

    MT output analysis

    Before editing begins, the MT output is reviewed as a whole to identify recurring error patterns — common mistranslations specific to this engine on this content type, systematic register errors, terminology gaps, and formatting issues. This prevents the same error from being corrected 40 times in isolation without recognising the pattern.

  3. 03

    Post-editing in CAT tool

    Editing is carried out segment by segment in Trados Studio, memoQ, Phrase, or the client's preferred platform, with the source and MT output visible simultaneously. Translation memory leverage is applied throughout, and the glossary is checked against every key term. For full post-editing, every segment is treated as requiring potential rewrite, not just error correction.

  4. 04

    QA review and consistency check

    A full-document read-through checks terminology, register, formatting carried over from the MT draft, and all client-specific requirements. QA tools built into the CAT environment are run and resolved.

  5. 05

    Delivery and TM update

    The completed file is returned in the format the agency requires — XLIFF, SDLXLIFF, MQXLIFF, DOCX, or direct commit to the TMS. The translation memory is updated with confirmed post-edited segments, and corrections can be flagged for engine retraining where the client runs adaptive NMT.

Quality

Why quality Indonesian MTPE requires domain depth

The output of any MTPE service is bounded by the post-editor's ability to detect and correct errors. For Indonesian, that means legal-translation-level proficiency combined with genuine domain knowledge — not fluency alone.

  1. 01

    Register & formality

    Indonesian operates across multiple registers — formal Bahasa Indonesia, everyday Indonesian, Javanese-influenced constructions, and domain-specific registers. MT engines routinely confuse them. A post-editor without native-level proficiency cannot reliably detect or correct these errors.

  2. 02

    Legal & financial terminology

    Legal and financial Indonesian contains terms that look equivalent to English but carry different scope under Indonesian law. Legal translation certification from Universitas Indonesia, in both directions, is what backs terminological accuracy in these domains, not general post-editing training.

  3. 03

    Technical precision

    In-house experience at Ford Motor Company and Toyota Motor Asia Pacific built a deep technical lexicon in automotive and engineering Indonesian — the specialised sub-vocabularies where MT engines typically perform worst.

  • Trados Studio
  • memoQ
  • Phrase (Memsource)
  • Wordfast
  • XTM Cloud
  • Smartcat
  • Lokalise
  • Crowdin

When to use MTPE

Indonesian MTPE services by content type

MTPE is not appropriate for every project type or quality threshold. These are the content categories where it delivers consistent value.

  • High-volume documents

    Regulatory filings, HR policies, operational manuals and product catalogues — where terminology and accuracy matter but volume makes translation from scratch cost-prohibitive.

    Recommended: full PE
  • E-commerce & product descriptions

    High-SKU catalogues and marketplace listings where NMT drafts are reasonable but brand voice and local consumer tone need professional post-editing.

    Recommended: full PE
  • Software & website localisation

    UI strings, help documentation and marketing pages, with terminology consistency, string-level issues and character limits handled by the post-editor.

    Recommended: full PE
  • Internal communications

    Policy documents, memos, HR communications and briefings for Indonesian-speaking staff where functional accuracy is enough.

    Recommended: light PE
  • Technical & user documentation

    Specifications, user manuals and technical guides where MT is accurate on standard terms but needs expert correction on specialised vocabulary and register.

    Recommended: full PE
  • Financial & compliance documents

    Financial reports, compliance summaries and regulatory disclosures where MT speeds production but financial and legal terminology must be exact.

    Recommended: full PE

Why this service

What a senior legal-certified translator brings to MTPE

The Indonesian MTPE market is dominated by general freelancers without domain credentials. For quality-sensitive workflows — legal, financial, pharmaceutical, compliance — the post-editor's qualification matters as much as the engine's output.

  1. 01

    Certified legal translation accuracy

    Dual legal translation certification from Universitas Indonesia means legal and financial terminology is checked by someone certified to translate it, in both directions.

  2. 02

    Deep domain specialisation

    Legal, financial, technical, and marketing domains — backed by in-house experience at the World Bank, Ford, Toyota, and UOB. Domain errors in MT output are caught because the terminology is actively known, not looked up.

  3. 03

    Agency-native workflow

    Trados, memoQ, Phrase, XTM, Smartcat — TM leverage honoured, client glossaries applied from segment one, files returned in the agency's required format.

  4. 04

    TM leverage discounts honoured

    Standard fuzzy band pricing is honoured. High-repetition batches and TM-heavy projects are priced accordingly; volume arrangements for ongoing workflows are available on request.

MTPE or human translation?

When to choose MTPE vs. human translation

ConsiderationMTPEHuman translation
High volume (10k+ words)Cost-effective with MT draftHigher cost per word
Publication-quality outputFull PE achieves thisStandard output
Sworn / certified requiredNot applicable to MTPECertified on request; sworn routed where required
Tight deadlineFaster with MT draftStandard turnaround
Domain: legal / financialFull PE by a legal-certified specialistHuman translation preferred
Recurring content typeTM leverage improves over timeConsistent, without MT savings
Court / official submissionNot appropriateHuman translation required
Agency NMT integrationDesigned for this workflowParallel track

Official submissions

If your project requires a sworn or certified translation — for a court, embassy, immigration authority, or government body — post-editing is not the appropriate service. See Legal Translation.

Common questions

Questions clients ask first

What is machine translation post-editing (MTPE)?

Machine translation post-editing is the process by which a professional translator reviews and corrects the output of a machine translation engine to meet a defined quality standard. It is covered by ISO 18587, the international standard for post-editing of machine translation output. MTPE is distinct from translation from scratch: the MT engine produces the initial draft, and the human post-editor corrects accuracy, fluency, terminology, and register. For Indonesian, the quality of MT output varies significantly by domain and text type, making the post-editor's expertise the primary quality determinant.

What is the difference between light post-editing (LPE) and full post-editing (FPE)?

ISO 18587 defines two levels. Light post-editing (LPE) targets functional accuracy and intelligibility — errors that change meaning are corrected, but stylistic improvement is not required. The output is understandable and factually correct, but may read as machine-translated. Full post-editing (FPE) targets human-translation standard — accurate, fluent, terminologically consistent, and appropriate in register. The intended use determines the appropriate level. Published, client-facing, or officially submitted content requires full PE. Internal reference or gist purposes suit light PE.

Which MT engines do you post-edit for Indonesian?

Post-editing services are available for output from DeepL, Google Translate (NMT), Microsoft Translator, ModernMT, Amazon Translate, SYSTRAN, Omniscien, and custom NMT engines integrated with Trados GroupShare or other TMS platforms. The engine does not change the post-editing process — errors are corrected against the source, regardless of where they originate.

How are MTPE rates structured?

Indonesian MTPE is priced per source word, with separate rates for light PE and full PE. The rate reflects the post-editing level, domain complexity, and MT engine quality. Translation memory leverage discounts are applied using standard fuzzy match bands — 100% matches and in-context exact matches are typically excluded or significantly discounted. For high-volume ongoing projects, a volume rate or retainer arrangement can be agreed. Contact directly with project details — engine used, content type, approximate word volume, and delivery cadence — for a specific rate proposal.

Can you work directly inside our TMS?

Yes. Direct login access is supported for Phrase (Memsource), XTM Cloud, Smartcat, Lokalise, Crowdin, and similar cloud-based TMS platforms. For Trados Studio and memoQ-based workflows, SDLXLIFF and MQXLIFF packages are accepted and returned in the same format. The preference is to work in whatever system your project managers already use — no workflow overhead on your end.

Is Indonesian MTPE appropriate for sworn or certified translation requirements?

No. Sworn and certified translation — required for submission to courts, embassies, immigration authorities, and Indonesian government bodies — must be produced as a human translation, not post-edited machine output. If your project or end client requires a sworn or certified translation, please see the Legal Translation or Certified Translation service pages.

What turnaround time can I expect for Indonesian MTPE projects?

Post-editing throughput for Indonesian is approximately 2,000–3,000 words per working day for full PE on domain-specific content, and higher for light PE or general content. Actual turnaround depends on MT output quality, domain complexity, and current project queue. For ongoing agency projects with regular delivery cadences, a standing turnaround schedule is agreed at project start. Rush delivery is available for urgent batches — mention your deadline upfront and feasibility will be confirmed with your quote.

Send a sample of the MT output.

Engine, level, volume and CAT tool are enough to quote. I reply personally, within one business day.

Send a project brief