Translation Memory Systems
The translation process can be aided by machine translation tools in a significant way. Translation memory systems (TMS), which compare a new source text to translations still on hand and identify to a certain extent agreements of sentences and formulations, are frequently used. Parts of sentences up to text blocks can be identified as "already translated" or "partially translated" and faithfully translated more quickly. TMS operate on the basis of terminology databases which are not restricted to single words and can search for agreements in more complex sentences too.
An important aspect for working with TMS efficiently is the consistent use of style sheets to increase the probably of hits and to keep post-processing to a minimum. Observing conventions in text presentation for the translator when working with TMS is also very helpful.
Translation Process Without TMS
The translator goes through the text step by step and translates smaller units of meaning. If a same or similar sentence appears again and the translator remembers that he translated this sentence once before, he will search for how he previously translated the sentence. The old translation will, as far as practicable, continue to be used. Unknown terms are looked up in word lists, glossaries or online lexica.
Translation Process with TMS
The TMS breaks up the text into smaller units. Segmentation rules are available to the system for this purpose. Since it is possible that not every unit represents a complete sentence, one frequently refers to segments here instead of sentences. For example, segments can be entire sentences, standalone words like "caution", short headings or even subsets in an enumeration list.
The TMS provides the translator with the segments for editing and stores the source segments and translations (segment pairs).
Each new segment to be translated is automatically checked against every previously edited segment in the TMS. If the system finds the same or a similar segment, it provides the translator with the translation of the old segment as a suggestion. The translator decides whether the translation is used, modified or regenerated.
Every term in the sentence being translated is looked up automatically in the TMS' terminology database. If a term is found, its translation is suggested to the translator.
The processes are, thus, very similar, with the only difference being that the TMS checks every new segment against all previously translated segments in any event and looks up each term automatically as well. This helps the translator to work more consistently and saves time when looking up old translations.
How Does the Translation Memory System Work?
The underlying idea behind every TMS is to have to translate a segment only once and to immediately provide a translation whenever the same segment turns up again. This is characterised as a 100% match (exact equivalent). If just a similar segment is found in the memory, this is frequently characterised as a fuzzy match (unsharp or similar equivalent). The work takes place in a bilingual environment, i.e. the translator always keeps both language pairs in view on the screen.
At the outset, though, the TMS is empty, i.e. the translator has to fill the fields of the database with his own translations himself. The terminology component is usually empty too, however there is the possibility of integrating commercial glossaries here in some cases. The segment pairs are stored in a database without context (without relation to surrounding text). The database is used for every translation involving this language pair.
Advantages of a Translation Memory
Translation memories provide many solid advantages over the course of the translation process:
- automatic reuse of already existing translations
- ensuring consistency
- consistent use of your preferred terminology
- reduction of translation costs by up to 70 %
- minimising costs for desktop publishing
- development of comprehensive customer-specific glossaries
Our Input
In the translation memory sphere, our experts and others will assist you in achieving the following:
- storing customer TMs
- restoring files for word analysis and translation with CAT tools
- reading out PDFs
- reading out images containing text / copying out images containing text
- protecting texts
- creating optimal data preparation workflows
- writing scripts for data preparation
- developing workarounds
- post-processing data / restoring translations
- revising existing TMs
- selecting the appropriate translation memory system
