Machine translation with human post-editing under ISO 18587

Machine translation with post-editing combines a first output generated by a machine translation system with subsequent human correction. The aim is to use language technology where it adds value, without delivering a text that has not been checked by a professional.

LinguaVox provides machine translation with post-editing for companies that need speed, multilingual capacity and professional control. The service is not the same as using an online tool and accepting the result. It includes feasibility assessment, preparation, human post-editing and final verification.

This workflow is linked to ISO 18587 when the project requires full post-editing by a qualified linguist of machine-translation output. It can be effective for technical documentation, software, product information, web content, training material and recurring multilingual content.

What machine translation with post-editing means

The process begins with machine translation. A system produces a first version in the target language. A professional post-editor then reviews that output against the source text and corrects it according to the agreed quality level.

The human part is not optional when the text must be used professionally. Machine translation can help create a base, but it does not guarantee meaning, terminology, style or formatting. It can produce plausible errors that are difficult to detect without bilingual review.

For this reason, the workflow must be planned. The client, provider and post-editor must know what type of text is being processed, what resources are available and what intended use the document will have.

Why raw machine translation is not enough

Raw machine translation can be effective for understanding a text quickly. It is not simply normally suitable for publication, client delivery or technical use without review. A fluent automatic output can still contain serious errors.

The tool may mistranslate a term, omit a sentence, change the tone, alter a number, damage a tag or ignore a client’s preferred terminology. These problems may be acceptable for informal reading, but not for professional documentation.

Human post-editing reduces that risk. The post-editor checks the output, corrects errors and decides when a segment must be rewritten. In a full ISO 18587 workflow, the goal is a final text comparable to human translation.

When to use machine translation with post-editing

Texts with volume and repetition

This workflow often works well when the content is large, repetitive and structured. Manuals, product sheets, help articles, technical instructions and knowledge base content may be good candidates.

Available language resources

Translation memories, glossaries and previous approved translations improve the process. They help control terminology and reduce variation across the final document.

Clear source text and suitable language pair

The original text must be clear enough for the machine translation engine to produce a usable base. The language pair also matters. Some combinations produce better output than others.

When we do not recommend this service

We do not recommend machine translation with post-editing for every document. Creative marketing, sensitive legal content, sworn translations, medical reports with high risk, contracts and texts with many ambiguities may require human translation.

It may also be a poor option when the source text is badly written or when the automatic output is not usable. In those cases, post-editing can become slower than translating from scratch.

A professional provider should not force machine translation into a project just to reduce the apparent price. The right workflow depends on the document.

How LinguaVox assesses project feasibility

Review of the source document

We analyse the source text, subject matter, language pair, volume, format and intended use. We also check whether the document includes tables, tags, variables or non-editable content.

Test of machine-translation output

When needed, we review a sample of machine-translation output. This helps estimate whether the post-editor will be correcting a useful base or rewriting most of the content.

Definition of scope

We define whether the project needs light post-editing, full post-editing, human translation or a mixed process. The scope must be clear before the quote is accepted.

ISO 18587 workflow for machine translation with post-editing

Pre-production: preparing before translation

The project is prepared with instructions, terminology, translation memories, files and expected quality level. If necessary, the source text is pre-edited.

Production: human post-editing

The post-editor compares source and target texts. They correct meaning, terminology, style, consistency and formatting. They also report recurring machine translation problems when relevant.

Post-production: final control and delivery

The final document is checked before delivery. This may include formatting, terminology, consistency and compliance with instructions.

Machine translation with post-editing and human translation

Human translation starts with a translator producing the target text. Machine translation with post-editing starts with an automatic output. The professional task is different, although both require linguistic competence.

Machine translation with post-editing can be efficient when the output is good and the content is suitable. Human translation is safer when the text requires creativity, interpretation, legal precision or deep adaptation.

The choice should not be ideological. The best workflow is the one that produces the required result with the appropriate level of risk and effort.

Typical sectors and document types

This service is often considered for technical manuals, maintenance documentation, user guides, product sheets, catalogues, software strings, online help, e-learning content, internal procedures, knowledge bases and web content.

It can also support multilingual updates when previous translations and terminology resources exist. In recurring projects, the combination of machine translation, memories and post-editing can improve consistency and turnaround time.

Each sector still requires specific control. A medical document, a software interface and a technical manual do not create the same risks.

Risks reduced by human post-editing

Human post-editing reduces the risk of meaning errors, terminology mistakes, omissions, additions, formatting issues and inconsistent style. It also helps identify whether the machine translation engine is producing recurring problems.

This is especially important when the text will be seen by clients, employees, users or regulators. A bad automatic translation can damage understanding and credibility.

The value of the service lies in combining technology with professional judgement. The machine creates a base. The post-editor decides what is acceptable.

Why hire an ISO 18587 certified company

An ISO 18587 certified company must work with a controlled process for full post-editing by a qualified linguist. This provides a clearer framework than sending documents through a tool and correcting them informally.

The certification does not mean that every document should be machine translated. It means that the provider has a process for evaluating, preparing, post-editing and checking suitable projects.

LinguaVox uses this approach to help clients decide when machine translation is effective and when another workflow is safer.

How to prepare a quote request

To request a quote, send the source document, target languages, intended use, deadline, file format and any glossary, translation memory or previous translation available. If you already have machine-translation output, include it.

We will review the material and tell you whether machine translation with post-editing is viable. If the project needs translation quality evaluation or another workflow, we will indicate it before quoting.

The more context we receive, the more accurate the recommendation will be.

Operational factors that affect the MTPE workflow

The success of machine translation with post-editing depends on more than the engine. The source document, file structure, segmentation, terminology resources, language pair and expected quality all influence the final effort. A good workflow starts before the machine translation stage.

If the source file contains inconsistent terminology, very long sentences, copied text, hidden formatting or unclear instructions, the automatic output will usually reproduce or amplify those problems. In these cases, pre-editing or document preparation can improve the result.

The post-editor also needs to know whether the final text is for publication, internal use, support, training, software, product documentation or regulatory communication. Each use creates different risk and style requirements.

Role of tools and engines

Different tools can produce different results. DeepL, Google Translate, Microsoft Translator, Amazon Translate, ModernMT, Systran and specialised tools may perform differently depending on the language pair and subject matter. The best choice is not always the most popular tool.

In professional projects, the engine is only one part of the workflow. Translation memories, glossaries, termbases, style guides and quality checks often have more impact on consistency than the engine itself. This is why the project must be prepared before the output is generated.

The same engine can work well for one type of content and poorly for another. A technical manual, a product catalogue and a marketing page do not pose the same problems. Testing a sample can prevent wrong assumptions.

Managing updates and recurring content

Machine translation with post-editing can be effective when content is updated regularly. If a previous translation exists and only some segments have changed, the work can combine translation memory leverage, machine translation and human post-editing.

This is common in technical documentation, software help, e-commerce catalogues and corporate knowledge bases. The objective is not just speed, but controlled reuse of approved language.

For recurring projects, LinguaVox can help maintain terminology and instructions across deliveries. This makes later projects more consistent and reduces the risk of repeating the same corrections.

Preparing the source text before machine translation

Machine translation with post-editing works better when the source text is prepared. This does not mean rewriting the whole document. It means removing avoidable ambiguity, fixing inconsistent terminology, checking incomplete sentences and clarifying passages that could be misread by the engine.

Pre-editing is especially useful when the same source document will be translated into several languages. A small improvement in the original can reduce errors across all target languages. It can also reduce the time spent by post-editors correcting the same problem again and again.

This step is not always necessary. It is more relevant in technical, procedural and multilingual projects where the source text contains long sentences, inconsistent wording or repeated instructions. LinguaVox can indicate whether pre-editing is worthwhile before the workflow starts.

Working with client-owned machine-translation output

Some clients already produce their own machine-translation output before contacting a language service provider. LinguaVox can work with that material, but it must be assessed first. The fact that an output exists does not mean that it is efficient to post-edit it.

If the file has been translated with a general tool, we check whether the terminology, sentence structure, formatting and completeness are acceptable. If the output is too weak, a new translation from the source text may be faster and safer. If it is usable, we can define the scope of post-editing and prepare a quote.

This assessment is important because poor machine translation can hide problems behind fluent language. A sentence may look finished but change the meaning. A term may be common but wrong for the field. A variable or tag may have been altered. These issues affect both quality and cost.

Multilingual MTPE project management

In multilingual projects, the same content may not behave equally in every target language. A machine translation engine may produce a strong output for one pair and a weak output for another. It may handle English to French well, but struggle with a smaller language or a highly specialised field.

For this reason, LinguaVox can apply different workflows within the same multilingual project. One language may use full post-editing, another may require human translation, and another may need additional terminology preparation. This is more realistic than forcing a single approach across all languages.

Project management also matters. Shared instructions, centralised terminology, consistent file preparation and feedback from post-editors help control the final result. Without coordination, multilingual MTPE can create inconsistent terms, uneven style and repeated corrections.

How feedback improves later projects

Post-editing produces useful information. If the same type of error appears repeatedly, it can indicate a problem with the engine, the glossary, the source text or the instructions. Recording that feedback helps improve future projects.

For example, a recurring mistranslation of a product component can be added to the glossary. A source sentence that causes problems in several languages can be rewritten. A formatting issue can be corrected before the next batch. This is one of the advantages of managing MTPE as a process rather than as a one-off correction.

Over time, this approach can make recurring multilingual projects more stable. The aim is not only to deliver one corrected document, but to improve the way similar documents are handled in the future.

Frequently asked questions about machine translation with post-editing

What does machine translation with post-editing mean?

It means using machine translation to create a first output and then having a professional post-editor correct it against the source text.

Is MTPE the same as post-editing?

MTPE usually refers to machine translation post-editing. Post-editing is the human correction stage within that workflow.

Can MTPE reach professional quality?

Yes, if the text is suitable and full post-editing by a qualified linguist is applied. It is not simply guaranteed by the machine translation engine alone.

Is it always cheaper than human translation?

No. It can be more efficient in suitable projects, but poor output may require as much effort as human translation.

Can I send machine translation already generated by my company?

Yes. We can assess whether the output is usable and quote post-editing if the quality is sufficient.

What does ISO 18587 add?

ISO 18587 provides a framework for full post-editing by a qualified linguist of machine-translation output, including process and professional competence.

Request a quote for machine translation with post-editing

Send us your files and project details. LinguaVox will assess whether machine translation with post-editing under ISO 18587 is the right option or whether human translation would be safer.