Some things that seem self-evident and hardly worth mentioning to us as a language service provider do deserve our attention. For example, when asked why there are two different systems for storing translations, i.e. the translation memory system and the terminological database. What is the actual difference?
Our answer is that both systems perform different tasks but work best when combined:
A translation memory system is a store for translations that assists translators by providing passages of text that have already been translated and can therefore be reused. To enable this processing, every piece of translated text is stored in the translation memory (TM) in segments. New texts, on the other hand, are split into individual segments before translation begins so that they can be compared with what is in the TM. A segment usually corresponds to a sentence, as the beginning and end of a segment is defined, among other things, by punctuation marks.
During the translation process, the TM system uses an algorithm to search for matches in the translation store for each new segment and then shows the relevant hits. The translator can then insert these suggestions and edit them as necessary.
In a terminological database however, only single words or very short phrases are stored. The database is in effect a dictionary that has the express purpose of providing detailed information such as technical terms and definitions. Other metadata can be stored in addition, such as “forbidden term” or contextual information including images.
Terminology is initially defined in the source language and subsequently added in the target languages. The idea is to have the greatest degree of control over the key terms used in a text and, as a consequence, to ensure consistency. Consistent terminology is an important factor for clear corporate language, to avoid misunderstandings and to ensure that the texts are easy to understand. The use of consistent and clear terminology allows companies to define their brand and stand out from the competition.
But the question still remains: why not store key terms and short phrases in a single system, in exactly the same way as longer segments are stored? Why are there two different systems for this?
The reason being that a TM is primarily based on quantity and the terminological database on quality. The TM is there to make use of previously translated texts so that these need not be retranslated or typed out each time.* The terminology, i.e. individual key terms in a text, is a qualitative definition that is intended to avoid misunderstandings about what is actually meant. Accordingly, both systems are structured differently.
In his work, the translator combines the two systems: on the one hand, the TM provides translation suggestions that can be edited. On the other hand, the terminological database – which is connected to the system – signals that there is a defined term in the segment to be translated. The translator does not need to research this term because it is shown to them as soon as it occurs in the text. A good translation memory system also enables an automatic terminology check from the terminological database as part of quality assurance.
Conclusion: TM systems and terminological databases have different tasks to perform and complement each other because the terminological database helps the translator to use defined terms consistently. To this end, the TM system helps to recycle sentences or phrases that have already been translated, which saves both time and money. Both systems in combination form an ideal working environment for translators to efficiently produce consistent and high-quality translations.
* The suggestions are based purely on character matching but do not confirm that the content is actually correct in context or which variant is preferable if there are several differing hits.