Use a custom glossary for more accurate machine translation

Machine translation can be impressively fast. But if you use tools such as DeepL, Google or ChatGPT regularly, you may also have noticed something frustrating: the same term is translated differently from one document to another, a product name suddenly changes, or the wording simply does not sound like your company.

The translation may be technically correct, but that does not necessarily make it your translation.

A custom glossary for machine translation gives translation tools clearer instructions about the terminology your organisation uses. By defining preferred translations, brand terminology and words that should remain unchanged, you can make automated translation considerably more consistent.

So, how does it work?

‍

What is a custom glossary in machine translation?

A custom glossary is a collection of approved terms and their preferred translations.

Think of it as a terminology rulebook for your translation tools. Rather than allowing the system to decide how a particular word should be translated every time it appears, the glossary tells it which terminology your organisation prefers.

For example:

‍

This becomes especially useful when the same content is translated across different teams, markets or platforms.

A glossary is just one of the tools used within a broader terminology management strategy, helping organisations define, organise and maintain their terminology consistently. 

Machine translation tools are designed to interpret language from context, but they do not automatically know how your organisation uses particular terms. A word may have several valid translations, and the most appropriate option may depend on your industry, product or audience.

A glossary removes some of that guesswork.

‍

How does a glossary improve machine translation?

A custom glossary creates a shared terminology reference that can support both automated and human translation workflows.

The main benefits include:

• Greater terminology consistency. The same concepts are described using the same approved terminology across documents and languages;
• More accurate specialised translations. Technical or industry-specific terms can be linked to the meaning that applies to your particular context;
• Better control over brand and product terminology. Product names, trademarks, features and abbreviations can be handled according to predefined rules;
• Less repetitive post-editing. Translators and reviewers spend less time correcting the same terminology again and again;
• Easier multilingual scaling. As translation volumes and language combinations grow, approved terminology provides a common reference for everyone involved.

However, a glossary does not control every aspect of a translation. It primarily manages terminology. Tone of voice, sentence structure, cultural nuance and creative messaging may still require localisation expertise, human review or a dedicated style guide.

‍

What should a machine translation glossary include?

The best glossaries focus on terminology where consistency genuinely matters.

Depending on the organisation, this might include:

• Company, brand and product names;
• Industry-specific or technical terminology;
• Product features and service names;
• Abbreviations and acronyms;
• Words that should remain untranslated;
• Approved translations for ambiguous terms;
• Legal or regulatory terminology;
• Terms that regularly create inconsistencies during translation.

The goal is not to include every word your organisation uses.

Common everyday vocabulary rarely needs terminology management. A useful glossary focuses on words and expressions where choosing the wrong translation could affect clarity, accuracy or brand consistency.

Context also matters. Adding a short definition, example or usage note can help clarify how a term should be understood, particularly when the same word can have several meanings.

‍

Glossary, translation memory and style guide: what’s the difference?

Glossaries, translation memories and style guides can all improve consistency, but they do different jobs.

‍

Used together, they provide much stronger guidance than machine translation alone.

A glossary keeps terminology consistent. A translation memory helps reuse previously approved content. A style guide helps ensure the resulting language still sounds like the organisation behind it.

‍

How can you use a custom glossary with machine translation and AI?

Glossary integration varies depending on the technology being used.

Platforms such as DeepL and Google Cloud Translation offer dedicated glossary functionality that allows organisations to define preferred translations for specific terms.

CAT tools such as Trados and memoQ can also incorporate terminology resources into professional translation workflows, giving linguists access to approved terminology while they work.

With ChatGPT and other large language models, the approach may be different. Terminology can be provided as part of the context or instructions given to the model, helping guide the output towards the preferred language.

The underlying principle remains the same: the translation system performs better when it has access to relevant information about how your organisation uses language.

Creating that resource usually starts with existing content. Websites, product documentation, marketing materials, technical manuals and previous translations can all contain useful terminology.

A practical process involves three main stages:

1. Identify the terminology that matters. Extract recurring or business-specific terms from existing content;
2. Review and validate it. Linguists and, where appropriate, subject-matter experts confirm which terms should be included and how they should be translated;
3. Integrate and maintain it. Add the glossary to the relevant translation workflow and update it as products, services and terminology evolve.

AI-assisted term extraction can make the initial process faster, but expert review helps ensure that the terminology selected is accurate and useful.

‍

When is a custom glossary worth using?

Not every translation project requires a dedicated glossary.

But terminology management becomes increasingly valuable when an organisation:

• Translates large volumes of recurring content;
• Works across several languages or markets;
• Uses specialised technical or industry terminology;
• Has a large product or service portfolio;
• Relies on several teams or translation tools;
• Combines human translation, machine translation and AI;
• Repeatedly corrects the same terms during review.

In these situations, inconsistent terminology can quickly become more than a linguistic detail. It can create extra review work, confuse customers and make multilingual content feel fragmented.

A well-managed glossary gives machine translation and AI systems clearer linguistic guidance, reducing unnecessary corrections and making multilingual workflows easier to scale.

Machine translation provides the speed. Terminology consistency gives that speed direction.

For organisations using MT and AI at scale, combining the right technology with well-managed terminology can make automated translation more accurate, more consistent and much easier to control.

‍