Professional machine translation with expert human review

Machine translation can be effective in multilingual projects with high volume, tight deadlines or repetitive content. But not every text is suitable for this workflow, and not every automatic output can be used without expert human review. In professional documentation, the real difference lies in knowing when to use it and when to reject it.

LinguaVox uses machine translation within controlled processes, not as an automatic substitute for professional translation. When the project allows it, we combine technology, linguistic resources and human post-editing to deliver a professionally reviewed result. When the text is not suitable, we recommend direct human translation.

ISO 18587 is the main reference for full post-editing by a qualified linguist of machine-translation output. If your company wants to use machine translation for content that will be published, delivered to clients or integrated into technical processes, the question is not only which engine to use. The question that matters is which level of human control the output needs.

What machine translation is

Machine translation is the automatic production of a target text from a source text by means of a computational system. Current systems are normally neural or AI-based and can produce fluent output in many language pairs.

That fluency can be effective, but it can also be misleading. A sentence may sound natural and still change the meaning, omit a condition, use the wrong term or introduce information that was not in the source. This is why machine translation must be assessed before it is used in a professional context.

Machine translation should therefore be understood as a possible starting point. It is not simply a final quality guarantee. The output needs a decision: accept it as informal information, send it for full post-editing, combine it with translation memories or replace it with human translation.

Machine translation for companies

Companies often consider machine translation when they have large volumes, frequent updates or multilingual content that must be delivered quickly. Product catalogues, help centres, knowledge bases, software documentation, internal procedures and technical updates are common examples.

The advantage is not only speed. In the right workflow, machine translation can help process repetitive material, support multilingual production and reduce manual effort on segments that are predictable. But this only works if the content is suitable and if the output is checked by professionals.

For companies, the operational question is practical: what will the text be used for, who will read it, what happens if it contains an error and what quality level is required? A public technical page does not have the same risk as an internal note. A legal clause does not have the same tolerance as a product description.

When machine translation is effective

Machine translation is usually more useful when the source text is clear, structured and repetitive. It may work well with technical documentation, product data, internal knowledge bases, support articles, software help, maintenance instructions and recurring corporate content.

It can also help in update workflows. If a previous translation exists and only some segments have changed, a project can combine translation memory, machine translation and human post-editing. This avoids translating everything again and allows previously approved language to be reused.

The best candidates normally share three characteristics: stable terminology, predictable sentence structure and a clear intended use. When these conditions are present, machine translation with post-editing can be a realistic option.

Common use cases

Common use cases include multilingual product catalogues, internal manuals, technical support content, software documentation, e-learning modules, corporate knowledge bases, recurrent instructions, product descriptions and documentation updates.

It can also be used in large projects where a first assessment shows that the output is strong enough to justify post-editing. In that case, LinguaVox can define a controlled workflow and assign suitable post-editors.

The important point is not the category alone. A technical manual may be suitable if it is clear and consistent. The same type of document may be unsuitable if the source text is ambiguous, poorly written or highly specialised without terminology resources.

When machine translation is not advisable

Machine translation should not be used automatically for every document. It is often unsuitable for sworn translations, sensitive contracts, legal filings, high-risk medical documents, creative marketing, brand voice content, investor communications, litigation material or documents where a mistranslation can create serious consequences.

It may also be unsuitable when the source text is unclear. If the original contains contradictions, incomplete sentences, badly scanned content, uncontrolled terminology or ambiguous instructions, the automatic output will often reproduce or amplify those problems.

In those cases, human translation is usually safer. LinguaVox can explain why machine translation is not recommended and propose another workflow: human translation, bilingual revision, terminology preparation, pre-editing or a mixed process.

Machine translation with post-editing

Machine translation with post-editing means that a human professional reviews and corrects the output of a machine translation system by comparing it with the source text. That does not mean the same as monolingual proofreading. The post-editor must check whether the automatic text says the same thing as the original.

The work may include correcting meaning errors, terminology, grammar, omissions, additions, numbers, units, tags, formatting, style and consistency. In full post-editing, the aim is to obtain a final text comparable to a professional human translation.

This workflow is effective only when the automatic output is a workable base. If too much of the text must be rewritten, the process stops being efficient. In that case, human translation may be clearer, safer and more predictable.

Full post-editing under ISO 18587

ISO 18587 focuses on full post-editing of machine-translation output. The goal is not to produce a rough text for internal orientation. The goal is to obtain a final output that is suitable for the agreed purpose and comparable to human translation output.

This requires qualified post-editors, project specifications, terminology resources and final checks. The post-editor must not only improve style. They must verify meaning and correct any problem that prevents the text from being accurate, complete and fit for its intended use.

For this reason, ISO 18587 is relevant when companies want to use machine translation in a professional setting. It provides a framework for deciding how the output should be treated and which human competences are needed.

Role of an ISO 18587 certified company

An ISO 18587 certified translation company helps the client decide whether machine translation is appropriate, prepares the project, assigns qualified post-editors and checks the final result. Certification does not mean that machine translation should be used in every project.

A certified company should be able to say no. If the source text, language pair, risk level or intended use makes the workflow unsuitable, the responsible recommendation is to use another process.

LinguaVox uses ISO 18587 as a quality framework, not as a marketing label. The process starts with an assessment of the content and ends with a checked delivery adapted to the client’s purpose.

Machine translation tools

There are many machine translation tools, including DeepL, Google Translate, Microsoft Translator, Amazon Translate, ModernMT, Systran and tools focused on specific languages. Some offer glossaries, APIs, file translation, integrations or enterprise functions.

The best tool depends on the language pair, field, confidentiality requirements, terminology resources and expected output. A tool that performs well in one project can be weak in another. It is not simply enough to choose the best-known engine.

In professional projects, the tool is only one part of the workflow. Translation memories, glossaries, style guides, source quality and human post-editing often have more impact on the final result than the choice of engine alone.

Pre-editing before machine translation

Pre-editing means preparing the source text before machine translation. It can include clarifying ambiguous sentences, correcting mistakes, simplifying overly long structures, standardising terminology and removing formatting problems.

This step is not always necessary, but it can be valuable in multilingual projects. A single ambiguous sentence in the source can cause different errors in several target languages. Correcting it before machine translation reduces the risk across the whole project.

LinguaVox may recommend pre-editing when the source text is likely to produce poor automatic output. This is especially relevant for technical documents, instructions, software content and recurring materials.

Linguistic resources: memories, glossaries and style guides

Machine translation works better when it is supported by linguistic resources. Translation memories help reuse previously approved translations. Glossaries define preferred terminology. Style guides indicate tone, conventions, formatting rules and choices that must remain consistent.

These resources do not eliminate the need for expert human review, but they reduce avoidable inconsistencies. They also help post-editors make decisions aligned with the client’s terminology and previous materials.

For recurring projects, the first assignment can be used to improve these resources. This makes later machine translation and post-editing projects more stable and efficient.

Confidentiality and use of tools

Confidentiality must be considered before using machine translation. Not every tool or plan offers the same level of data control, storage policy, access management or enterprise guarantees. Public web tools may not be appropriate for confidential documents.

A professional workflow should define what tool can be used, under which conditions and for which type of content. Contracts, personal data, unpublished product information, financial documents and internal technical material require particular caution.

LinguaVox assesses the confidentiality requirements before recommending machine translation. When the content is sensitive, we may advise against using general tools or propose a more controlled workflow.

Difference between machine translation and human translation

Machine translation produces text automatically. Human translation is produced by a professional translator who interprets the source text and creates the target text from the start. Post-editing is a third workflow: it starts from machine-translation output but requires human intervention.

The three options are not interchangeable. Machine translation can be effective for a first understanding. Human translation is often safer for sensitive, creative or high-risk documents. Post-editing can be efficient when the automatic output is strong enough and the final text must be professional.

The right choice depends on risk, purpose, language pair, content type, terminology and budget. A serious provider should explain the difference and recommend the workflow that fits the document, not the workflow that sounds cheapest.

Frequently asked questions about machine translation

What is machine translation?

Machine translation is the automatic production of a target text by a software system from a source text. It can be effective, but it does not guarantee professional quality by itself.

What is the difference between machine translation and post-editing?

Machine translation generates the text automatically. Post-editing is the human correction of that output by comparing it with the source text.

Is machine translation with post-editing cheaper?

It can be more efficient when the output is usable. If the output is poor and must be rewritten, human translation may be more efficient.

Which texts should not be machine translated?

Sensitive legal documents, sworn translations, high-risk medical texts, creative marketing and documents with serious consequences if mistranslated often require human translation.

Can LinguaVox review a machine translation already produced?

Yes. We can assess whether the output is usable and recommend full post-editing, human translation or another workflow.

Request a machine translation quote with expert human review

Send the source files, target languages, intended use and deadline. LinguaVox will assess whether machine translation with human post-editing is suitable for your project.