What is Translation Memory (TM)
This article introduces you to the concept of a Translation Memory, how it works and the many benefits it offers to you.
What is Translation Memory
Translation Memory (TM) is a linguistic database that serves as a central hub for all your human-made translations. It does this by continuously accumulating sentences translated for specific language pairs for current and future project reuse. TM is an integral cornerstone of any modern translation-related technology.
It should be made clear that TM is in no shape or form a Machine Translation. While Machine Translation focuses on automatic translation without any human input, TM is rather a universal method of reusing previous human-made translations for specific language pairs.
A TM database stores sections of text from the source and their translated counterparts as "segment pairs", also known as "translation units". Segments in a text are most commonly sentences, lists, tables, and number combinations. Text segments are defined by markers, which can be full stops at the end of a sentence or, for example, paragraph markers.
How TM Works
At the initial stage, a computer-assisted translation tool uses the TM to analyze and compare the current source file with previous TM files: if no TM file exists, a blank TM will be created that will serve as a central hub for all current and future projects for that specific language pair.
During the analysis of a document, TM compares new text segments with existing TM segments for a specific language pair and calculates how many times those segments repeat themselves in a new project.
Upon completion of the analysis, TM identifies any matching segments and separates them into four categories based on a percentage match. The linguistic team working on a specific project has to make a decision whether to fully or partially use the suggested segments or to ignore the suggested segment altogether and create a new translation. The match types are as follows:
Full match (100% match) – a segment that is fully identical to an existing TM segment.
Repetition – a segment that exists multiple times in a current document. It has to be translated only once and any future occurrence will be provided by TM. Repetitions are very common in new TM projects.
Fuzzy match (75-99% match) – a group of segments that are partially identical to an existing TM segment. In this case, only the non-identical parts of the segment have to be fully translated.
No matches (0-74% match) – a unique segment that does not repeat itself within the text nor has any matches in the TM and therefore has to be fully translated.
Translation Memory creation process
Translation Memory is created in two ways:
1) New TM – new translation memory is created during your first project and is continuously reused and further populated by current and future projects. This is the most common type of TM as it is created when no previous translation for a specific language pair has been done.
2) TM from existing translated documents – this type of TM is created when there is an existing translation of a source document into a target language. In these cases, the linguistic team initially carries out a process called "alignment" that populates an existing or new translation memory with the source text and its target language counterpart for use in future projects.
Benefits of using Translation Memory
Translation Memory can become a key asset to any business or organisation in both the short-term and long-term by continuously providing you with benefits and long-term dividends. Key benefits of Translation Memory are:
Cost reduction – Once TM has been created it serves you by continuously reducing your long-term costs since every repeating or matching segment that already exists in TM will become cheaper for you. As a result, only new unique segments have to be fully translated, while fuzzy-match segments have to be translated only partially. Large TMs are long-term major assets that work in your favour by continuously driving costs down.
Exact pricing – TM's ability to track all past and current projects simultaneously allow for exact quotes to be produced for new projects, allowing clients to know exactly how much they will have to pay a Language Service Provider (LSP).
Consistency – The ability to track changes in all the source texts improves the consistency of all of their translated variants. TM keeps specific terminology and phrasing consistent regardless of which linguist team is working on your project.
Universality – TM is universal and works regardless of the content type you use: document translation, website localisation, software localisation, game scripts, etc. TM can be used for any of these types of content and various others. For example, if your first translation project was a software localisation, your existing TM can be applied to a website localisation project.
Speed – TM greatly improves the speed of your translation project without impacting the quality by fully or partially reutilising previously translated segments based on the segment match types described under the heading “How TM Works”.
In every modern translation and localisation environment, TM stands as a solid core that just has to be used at all times thanks to the many benefits it provides, both for the linguist teams involved and for the client.
By utilisation of previously translated content, TM has a positive impact on the quality of any current and future project by optimising the workflow of the linguistic teams, decreasing the chance of mistakes, increasing translation speed, and continuously pushing down the cost for the client.
Usage of TM, both short-term and long-term, pushes the quality of any project onto a whole new level across the board.