Translation memory and machine translation have similar acronyms (Machine Translation Translation Memory), which can be confusing. However, they have significant differences. Translation memories are created over time as translators add new content to the database.
The translator uploads the source text to a translation management system. The system analyzes the source text and finds the corresponding translated fragment in the translation memory. The translator can view both the source text extracted from the database and the proposed translation. By comparing them with the source text, the translator can proceed to one of the following actions: accept the proposed translation, make changes to it, or create a completely new translation.
Translation memory software has a rating system to determine the accuracy of translation matches. If there are no matches, the translator needs to perform a new translation of the fragment, which will be added to memory and will be used in the future for other translations.
Machine translation provides text generated without any human intervention.
Machine translation usually faces the dilemma of quality and preserving the meaning of the source text. The machine can quickly process significant amounts of text. The problem is that machines and artificial intelligence do not have the same sense of language as humans. This is why post-editing by human translators is essential when using machine translation.
Using a translation memory is useful when you regularly need to translate content from one language to another, especially if all or part of that content has repetitive parts. However, for creative work, translation of advertising and artistic texts, translation memory does not give such good results as for technical and financial texts. However, even marketing translation can benefit from the use of machine memory over time. With a critical approach from the translator, terms can be translated faster for a specific target market.
