A translation memory (TM) is a database in which translated sentence or segment pairs have already been stored. Each stored sentence has a unit, the so-called translation unit, as a finished translation. For subsequent projects, the system compares new texts with these units and immediately suggests identicalor similar passages. In this way, translators can make targeted use of tried and tested formulations, thereby maintaining a high degree of terminological consistency and shortening their processing time.

In the following article, you can find out why the use of translation memory has many advantages, what exactly it is and how translation memory management works.

The importance of TM for localization

Translation memory is at the heart of modern localization processes. Why? Because the database consistently exploits linguistic repetitions. Instead of translating sections from scratch again and again, TM reactivates existing solutions. An approach that significantly increases quality, reduces costs and ensures adherence to deadlines without having to rely on automated processes.

How does translation memory work?

When adocument is opened, the translation software with translation memory breaks down the text into small units. Usually in individual sentences. Each new unitis automatically compared with the existing translation memory:

‍

• Matched sentences: If a sentence matches a saved variant completely, the matching translation appears immediately.
• Similar sentences: If the sentence differs only slightly, for example by a different date or an additional word, Translation-Memory points this out by indicating the similarity value, for example 87%. Deviations are highlighted in color sothat only these passages need to be adapted.

‍

After confirmation, the new record pair is written directly to the memory. As a result, the database grows with each project and delivers even more hits in future. Regardless of whether the source text comesfrom word processing files, websites or layout programs.

Several studies have shown that translators work significantly faster with translation memory without any loss of quality. On average, productivity increased by around a third. Ideal for projects with many repetitions or regular updates.

‍

How the intelligent translation memory from locize works

The Translation Memory (TM) from locize is an intelligent translation support system that runs directly in the browser. As a result, it offers particularly high performance and fast results, as no requests need to be sent to the server. In contrast to traditional translation memories, the locize solution allows multiple projects to be used as sources at the same time. This means that existing translations from different projects can be reused, which saves time and increases consistency.

‍

A special feature is that the search is not limited to exact language pairs. Similar language variants - such as British and American English (en-GB/en-US) - are also taken into account. Each hit in the TM is linked to the original project, which facilitates traceability and enables uniform translation across multiple projects.

‍

Configuration is very easy in the project settings under the “EDITOR, TM/MT, ORDERING” tab in the “Translation Memories” section of the “Cat settings”. There, users can specify which projects are to serve as the source for the translation memory. This service is activated by default in new projects. When working with a segment, a corresponding tab appears on the right-hand side of the editor where you can view and use the TM suggestions.

‍

‍

Learn more about our features and how to use it in our documentation.

‍

The advantages of translation memory

Anyone who frequently translates similar content into several languages will benefit in particular from translation memory. The database turns previous translations into an immediately usable resource and thus provides several levers for efficiency and quality.

‍

• Noticeably faster work: Identical matches appear automatically, allowing translators to concentrate on genuinely new passages.
• Significantly lower costs: reused segments do not need to be translated or checked. This in turn reduces the cost of fees and proofreading.
• Seamless consistency: If every recurring word paircomes from the same source, terminology, style and tone remain consistent across all projects. Even with multiple language service providers.
• Sustainable quality improvement: A curated TM only saves checked formulations, which reduces the error rate. Brand specifications can also be adhered to automatically.
• Centrally available knowledge base: everyone involved has access to the same database, regardlessof whether it is marketing, legal or technology. This prevents data silos and duplication of work.
• Shorter time-to-market: Less translation and post-editing time speeds up releases,product launches and campaign launches.
• Direct sales leverage: High-quality, consistent localization increases the confidence of international customers and demonstrably increases the willingness to buy.
• Globally scalable:Modern databases cover hundreds of thousands of language pairs, opening thedoor to additional markets without having to set up new processes.
• Focus on creativity:Because routine passages are processed automatically, there is more capacityfor stylistic subtleties, cultural adaptations and complex content.

Translation memory: turbo for more speed, consistency and scope

‍

A professional translation memory turns every translated line in to capital. It speeds up projects, reduces costs, keeps terminology stable and shifts the focus away from routine copy & paste and towards stylistic and cultural fine-tuning.

Extra tip: Establish clear rules for administration, organization and maintenance from the outset, for example, only include approved segments in TM,regularly clean up duplicates and assign uniform meta data (topic, customer,version). This keeps the memory lean, reliable and valuable in the long term.

FAQ: Translation memory

1. How do you set up a translation memory?

• Collect inventory: Collect all existing bilingual files (e.g. previous Word translations, two-column Excel tables or bilingual InDesign content).
• Align: Use alignment tools to link source and target segments so that they flow into the translation memory as correct pairs.
• Import: Load the aligned TUs into your CAT solution (TMX or SDLTM format).
• Maintain: Define roles that release new segments, delete obsolete ones and merge duplicates.
• Versioning & backing up: Create backups and carry out major clean-ups in a copy toavoid data loss.
  
  

2. What is the difference between a translation memoryand a termbase?

A translation memory stores complete sentence or segment pairs (source and target text) and helps to reuse entire formulations. A termbase (terminology database) manages individual technical terms, often with definition, context sentence, grammar information and release status.

In short: TM = complete sentences for consistency and speed. Termbase = technical terms for precision and brand language.

3. what is a translation memory in RWS?

RWS is the manufacturer of Trados Studio (formerly SDLTrados). In the RWS ecosystem, the translation memory is an SDLTM file (local)or a cloud-based “translation memory” in Trados Live Team/Enterprise. The functional logic remains the same: segments are matched, hits are displayed and saved after approval. RWS TMs also support context matches (“CMs”) - they take segment adjacencies into account and therefore provide more precise reuse suggestions.

4. Which formats can a modern TM process?          

Depending on the CAT tool, typically: Microsoft Office, HTML, XML, XLIFF, JSON, Adobe InDesign/FrameMaker, Markdown, PO filesand many more. TM neutrality is important. The memory itself remainsfile-independent.

‍

5. Can I manage several languages in a single TM?

Yes, but only in bilingual pairs. For three target languages you need three TMs (DE-EN, DE-FR, DE-IT). However,multi-lingual databases can be searched together in many tools.

‍

6. How do I keep the quality of my TM high?

• Establish approval process (dual control principle)
• Run duplicate and error reports regularly
• Archive outdated segments with a time stamp
• Use standardized style and terminology guides

‍

7. What happens if the source contains small changes(e.g. new number))?        

The translation memory provides a match. You simply adjust the deviation, save the segment - and the updated wording is ready for the next similar passage.