Translation Memory & Terminology Management

A translation memory is a linguistic database that stores source sentences and their corresponding translations as segment pairs for future use. Usually a segment is a sentence, but it could also be just a few words, like a heading. If an identical or similar segment comes up later, it doesn’t need to be translated from scratch. Instead, the translation can be retrieved from the translation memory and, once checked for suitability in context, reused. Translation memory can be used to improve consistency, shorten turnaround times and reduce costs.

CAT stands for Computer-Assisted Translation (or Computer-Aided Translation) – a piece of software or web-based platform that provides the right environment and resources to support translators, such as translation memories, glossaries (termbases) and quality assurance tools. Some CAT tools also integrate with machine translation solutions and LLMs, allowing translators to use AI-generated suggestions if needed and then post-edit the output directly within their CAT tool.

Ideally, all translated content should be stored in translation memory databases, as you never know when you might need to retrieve it for future projects. Large volumes of work that contain repetitions are an ideal candidate, as translation memory will definitely help with consistency, turnaround times and cost. Content that’s regularly updated or reused across various projects is another great candidate. Highly creative, short texts you’re unlikely to reuse can still be processed using translation memory, but the advantages won’t be as significant due to the lack of repetition.

An exact match means a segment was translated before with absolutely no differences from the segment in the content we’re translating now. This is very useful, but it doesn’t automatically mean it’s the right translation for your new context. At Planet Languages, we always check exact matches for in-context accuracy and text coherence. Whenever similar, but not identical, matches are identified, these are called fuzzy matches. They generally range from 50% to 99% in similarity to previously translated content; anything under 75% is often considered entirely new text, as the similarity is too low to be of much relevance.

Upon receipt of your project, we analyse the content against your translation memories and provide you with a discount figure based on the translation memory report. Content repetitions and various matches – 100%, 95–99%, 85–94%, 75–84% and under 75% – are charged at different fractions of the normal word rate, based on their similarity to content already in the translation memory. For example, 95–99% matches are easier to amend than 75–84% matches, so the rate is lower.

A translation memory is a linguistic database that stores source sentences and their corresponding translations as segment pairs for future use. Within a CAT tool, a translator can look up similar matches to help maintain consistency with previously translated content and increase productivity. A machine translation system, in contrast, automatically translates text from one language to another. The linguist then post-edits the output, ensuring quality is in line with your requirements.

Of course – we can prepare a glossary based on the content we translate for you. Glossaries can vary in detail, from a simple source term/target term pairing to definitions, synonyms, who approved the content and when, product images and usage notes.

Our translation environment integrates with your glossaries and allows us to perform detailed quality checks, ensuring the terms in your translation match your approved terminology.

Planet Languages works with server-based glossaries and web-based platforms, so you can access your glossaries at any time. Access is controlled with different rights – we generally advise giving stakeholders read-only access, with editing rights reserved for experienced language professionals who can ensure any changes comply with your requirements.