The use of translation memory

The translator first provides the source text (that is, the text to be translated) to the translation memory. The program will first analyze this text and try to find out whether the existing translation part matches the previously translated text in the database. If a matching translation pair is found, it will be submitted to the translator for review. Translators can choose to accept, reject or modify the old translation. If it is modified, the modified version will also be recorded and stored in the database.

Some translation memory systems only search for words with a matching rate of 100%, which means that only new source words will be accurately compared with the database, and only exactly matched data will be displayed. Other systems use fuzzy comparison principle to find similar fragments and present them to translators with special tags to make them easy to identify. It is very important that the universal translation memory system will only search the source language from its database.

Completely mismatched text segments must be translated manually by translators. These newly translated text fragments will be stored in the database, and future translations may be adopted immediately due to the repeated occurrence of the source text.

Translation memory works well when the repetition rate of words in articles (such as some technical documents or manuals) is high. Translation memory is also very helpful to translate the situation that the existing documents have been modified step by step in the past. Generally speaking, translation memory is not considered in literary or creative documents, mainly because these types of articles are quite repetitive. However, some people think that these words with extremely low repetition rate are still worth collecting. For example, they can be used to search harmony. Other collected translation memories can also help with quality verification and proofreading (using translation memories, the source language and target language can be easily exported and paired in each line, and the source language and target language are separated by tabs, which is convenient for confirming whether there are untranslated words or for translators to proofread).

When the translation memory continuously uses appropriate words for a period of time, it will save a lot of work for translators.