tanslation memory is like personal machine translation Translation memory (TM) and machine translation (MT) are not that different. Of course there are some remarkable technical differences but the reality is that users don’t see the technology; they see what they can do with it. The basic concept behind both TM and MT are surprisingly similar and they can be used for same purposes.

A typical translation memory contains words, terms, expressions and other information that is relevant to the project. This information can be stored from previous translation projects. The benefits of having a TM are clear: increased consistency in terminology, high translation quality and faster translation process. A TM is often created if there are many similar documents to translate, or if the company wants to control the terminology, or if a translator has a long term translation relationship with the company. Some translators even create TMs for every new project and customer.

A typical machine translation contains a huge amount of language and translation specific data.  The purpose of all this material is to teach the machine to translate any piece of text. There are different types of MT: Rule Based MT works with translation and language rules. Statistical MT calculates probabilities and densities from a huge database. Hybrid MT combines both RBMT and SMT. Customized MT is tuned for certain types of texts. These differences are mostly technical, and all of them aim for creating automatic translations.

Of course there are also some differences. Translation memory is often project or company specific and managed by the translator, while machine translator is often created for wider use without full control of the output. TMs can be anything from very small to extremely large depending on the translation project. To be usable, MT requires a huge amount of information. TM ensures the correctness of terminology whereas MT is trying to translate the whole text.

So both TM and MT are used for automating translation work. Both contain information about the languages and translation rules. Only the scale of usage is different. Based on all this, it isn’t exaggerated to say that a TM is like a personal or project specific MT.


mlz-logoTake advantage of both MT and TM when localizing. Try Multilizer localization tools for software, documents and content localization. Use translation memories and/or automatic translations to ensure highest quality and most efficient localization process. Learn more.