Translation memories are an inherent part of the translation process of our time. The purpose of the translation memory technology is management of source texts and their respective translated segments in multiple languages. This means that previously translated text segments are re-used, thus contributing to decrease of cost of translation as well as ensuring consistency of translation texts and application of pre-defined language rules. Nowadays, it is common for a translation services provider to manage translation memories of more than 20,000 segments in 15 to 25 language combinations. Which is more than a few!
Which situations or issues must be considered?
A translation memory is used over the course of several years by multiple users which can lead to several translation variants of the same text. The translation memory could thus contain sentences with identical content but different variants of expression. In this case, the translator has to choose one of the translation variants which leads to inconsistency of the translation memory.
Moreover, the translator may be suggested outdated spelling or spelling mistakes which need to be corrected and confirmed. This only leads to further inconsistency of the translated segments contained in the translation memory and to the danger of the incorrect or outdated translation texts reappearing again in the future as suggestions.
The most common translation mistakes, such as terminology errors, typographical errors or numerical errors, can be immediately removed thanks to CAT tools and spell check tools. However, mistranslations, outdated texts or segmentation mistakes must be always considered by an individual.
To achieve a high quality of translation, it is necessary that translation texts in the translation memory are linguistically clean and occurrence of too many translation variants is avoided. In spite of all efforts, it is not uncommon to discover flaws in a translation memory even after several years. This leads us to the question:
Can the translation memory be cleaned of unwanted translations?
The translator or the TM creator can immediately correct or delete incorrect or duplicated translations and their variants. The translation memory can thus be cleaned of unwanted translations – instantly and regularly. However, if the translation memory has been used for years by multiple translators, this option is not so efficient any more.
The second option requires more effort and costs but the results can be observed much quicker. We are able to carry out a complex examination of the translation memory by means of CAT tools and clean it of doublets and errors. To identify them faster, we can use intelligent spell check tools. However, the individual variants must be reviewed and confirmed by a competent person (translator, terminology specialist, reviewer, manager).
A translation memory which has been cleaned in this manner is most efficient for future translations. It can significantly speed up the process of translation as well as checking and reviewing the translation. Nowadays, more and more companies decide to invest their time and finance in cleaning up their translation memories, preferring quality and consistency of future translations to the danger of incorrect translations.
The translations service provider must regularly assess the status of translation memory, inform the customer in timely manner and also manage the whole process of cleaning the translation memory.