Reuse validated segments · Maintain consistency across the lifecycle · Grounded AI, not guesswork.

Translation Memory Systems

The validated memory for consistent specialist and AI-assisted translations

Trados · ECQA · LICS · SDL · Quanos · tekom · ISO 9001

A translation does not have to start from scratch every time. A Translation Memory stores every validated source–target sentence pair and makes it available again for the next translation project. Recurring and similar passages are not translated from scratch. Instead, they are taken from the validated repository. This ensures consistency, faster turnaround and traceability over many years.

With machine translation and AI, this memory is becoming increasingly important. AI systems generate fluent text, but have no memory of previous decisions: They translate every variant afresh and may sound different after each update. The translation memory provides the memory that AI is missing. It provides the validated specifications that guide the engine and makes the output repeatable and auditable.

Glossary, terminology and translation memory together form the control layer that enables reliable translations: The glossary and termbase govern the terminology, while the translation memory governs complete segments.

What a translation memory can do

A translation memory delivers value in three key ways: by saving costs and time, by ensuring consistency across the entire lifecycle, and by serving as a foundation for controlled AI.

Reuse (costs & time)

Recurring sentences and updates are retrieved from the already validated content. When changes are made, only the new content is translated. The approved content remains exactly as it is. This reduces effort and turnaround times, especially for documentation with many versions.

Consistency across the lifecycle

Systems and products are operated and documented for more than 20 years. The translation memory keeps the language consistent across releases, variants and suppliers. Version 12 speaks the same language as Version 3.

Grounding & steering AI

The translation memory provides verified reference material that guides the machine translation. Instead of formulating freely, the engine applies tried-and-tested solutions. The output becomes repeatable and verifiable.

Request a project & get advice

Do you have technical documents that need translating? Send us 1–2 sample files. You will receive a specific recommendation and a quote within 24 hours.

Translation Memory and AI

Even with powerful AI, the translation memory does not become obsolete. It becomes even more important. The engine is rented and interchangeable. The translation memory, however, is your competitive edge.

Memory, not reinvention

AI has no memory of previous translations. The translation memory takes care of it. Reviewed segments take priority over freely generated suggestions.

Foundation for adaptive MT

Using your existing content, machine translation can be tailored to your language and solutions. The output becomes more accurate and requires less correction.

Control & audit trail

Every segment is human-reviewed and has a clearly documented source. In regulated industries, this is the foundation for ensuring that translations stand up to audits. The responsibility remains with the people (Human-in-the-loop).

Data sovereignty

Your language data is managed exclusively for you. It is kept separately and hosted on EU servers, so AI use never means losing control of your language data.

Find out how we steer the engine within the process on our page for AI-assisted Translation.

Data Protection & Data Security

Processing is GDPR-compliant. The servers of our translation engine are located in the EU. Your texts are used solely for your translations and are not used to train the engine. For each client, we set up a separate, company-specific data set: a dedicated Translation Memory, dedicated terminology and dedicated data. Our IT systems are protected against unauthorised access and are continuously maintained. You can find more details at IT Security and Data Protection. This ensures that our AI-supported translation processes operate in accordance with the EU AI Act: transparent, compliant with data protection requirements and with clearly assigned data.

Translation Memory, Termbase and Glossary in combination

Three building blocks interlock, from harmonised terminology through to the reviewed sentence:

Glossary

The harmonised list of terms: the low-threshold entry point to consistent terminology. Publicly available in our Glossary.

Termbase

Stores the binding terms together with their definitions and contexts, including permitted and prohibited usages for each language and subject area. Learn more on the page Terminology Database.

Translation Memory

Stores reviewed translation segments for reuse in complete sentences rather than individual terms.

In short: The termbase manages individual terms, whereas the translation memory manages complete sentences. The Language Operator connects these layers within the process: Technology, terminology and review interlock, while the responsibility for quality always lies with people.

Managed exclusively for your company

For every client we manage a separate, company-specific linguistic data set: a dedicated translation memory, dedicated terminology and dedicated data. This data set is used exclusively for your texts, handled in strict confidence and never mixed with other clients’ data.

We build this data set and maintain it continuously, ensuring it is cleaned, kept consistent and up to date across product generations. This ensures that your language stays reliable and auditable over many years, without you having to invest any effort in administration or maintenance yourself.

Processing is carried out in a GDPR-compliant manner on servers located in the EU. If you wish to integrate the data set into your own processes, for example into an authoring or editorial system, we will work with you to find the right solution. More on this: IT Security and Data Protection.

When a translation memory offers the greatest impact

Especially valuable for

recurring documentation with updates and variants (manuals, operating instructions, catalogues)
long product and system life cycles with ongoing maintenance
many languages, releases and suppliers, with consistency across multiple sources
use of AI/machine translation, where the translation memory grounds and steers the output
regulated content where traceability is essential (liability, audit)

Less essential for

one-off small projects without repetition
purely creative standalone texts where a fresh rephrasing is desired (transcreation)

Typical fields: Plant Engineering and Mechanical Engineering. As a service: Technical Translations.

Tools & Integration

We work as an RWS-certified partner with established CAT tools and TM systems, first and foremost Trados Studio with the MultiTerm terminology solution. We integrate your translation memory directly into the translation and QA process. With the GroupShare server solution, handover and revision take place directly, without detours via email.

Common input formats: Word, Excel, InDesign/IDML, FrameMaker/MIF, XML/DITA, CCMS exports and UI strings are processed segment by segment so that layout and structure are preserved. When multiple suppliers are involved, a shared translation memory keeps the language consistent throughout.

AI-assisted Translation How we steer engines and carry out post-editing in line with ISO 18587. Terminology Database Binding terminology as a control layer. Technical Translations The service where the translation memory is applied. Glossary Key terms relating to translation, technology and standards.

What is a translation memory?

A translation memory stores the reviewed translation segments (sentence pairs from the source and target text) and makes them available again for later projects. This keeps translations consistent and avoids having to recreate them from scratch. Find out more in the glossary: Translation Memory.

Do I still need a translation memory when AI is available?

More than ever. AI systems generate fluent text, but have no memory of previous decisions. The translation memory is that memory: It grounds the engine, ensures consistency across updates and makes the output auditable. Die Engine ist austauschbar — die geprüften Segmente sind es nicht. Genau deshalb führen wir für Sie einen eigenen, kundenspezifisch getrennten Bestand.

How do translation memory, termbase and glossary differ from each other?

The glossary is the agreed list of terms, the termbase holds binding designations with attributes, and the translation memory stores validated translation segments. In short: Term bank = terms, translation memory = full sentences. They complement each other; the definitions are in the glossary.

What are 100% matches and fuzzy matches?

A 100% match is a segment that has already been translated and reviewed exactly in the same way before. A fuzzy match is similar to a previous segment and is adjusted based on the suggested translation. This means that only new or modified content is actually translated from scratch.

Does a translation memory really save money?

For recurring documentation and regular updates, yes: Repetitions and matches are billed at a lower rate. Only the modified content is billed at the full rate. In addition, fewer review cycles are needed because the language remains consistent.

How are my translation memory and my linguistic data handled?

Your linguistic data is client-specific: For each client, we maintain a separate, company-specific repository that is used exclusively for your texts and treated in strict confidence, in a GDPR-compliant manner on an EU server. We create and continuously maintain the translation memory for you. This ongoing maintenance and administration effort is where the true added value lies. If you want to integrate the inventory into your own systems, we will coordinate how it is provided with you on an individual basis.

How does the translation memory stay consistent over the years?

Through continuous maintenance with clearly defined roles, quality assurance in accordance with ISO 17100/18587, and integration with the terminology. This ensures that the inventory, releases and variants are coherent throughout the entire lifecycle.

A memory that keeps your translations coherent

Send us 1–2 sample files. You'll receive a concrete assessment of reuse potential, effort and terminology setup. Confidential and without detours.

Book a Consultation Free Quote

ISO 17100 · ISO 18587 · ISO 9001: audited processes, clearly assigned data