How does Translation Memory work?

We will take a look at TM and explain to you how it works and its benefits. Ready to learn?

What is Translation Memory?

A simple way to explain it, TM or ‘Translation Memory’ is a database for a previously translated text that can be reused when needed. Another way of putting it, it’s a tool that recycles text and stores it for reuse.

How does it work?

Let’s start from the beginning, here… asking the question if you know which is the smallest unit in machine translation?
You know that a binary system has 0 and 1, and the combinations of these can make countless information. Language has letters that form words. They combine in sentences and afterwards in paragraphs and so on until they form the intended message.

what is translation memoryWhen it comes to Translation Memory, its building unit is a Segment. And these segments can vary from sentences to paragraphs, headings, titles and many others that have been translated previously.

Once you have the segments ready, a software called Translation Memory Manager (TMM) operates with it. Usually, they are a part of a CAT tool or any other form of translation assisting tool. That includes a translation management system (TMS) or even machine translation without any editing afterwards.

Behind the scenes, you can imagine a small robot running through the text. It compares it to the Translation Memory and says if there are matches. The matches can be on different levels: 100%, 70% and etc.

Depending on the type of Translation Memory, it can also compare the previous and next segments in the text and add context matches, which help the linguist choose a proper version of the translated segment. Additionally, the TM can be improved by adding more versions depending on the context.

You may also like:  Cybersecurity Translation in the Asian Context

What is our experience with TM?

The most important thing for a good translation is the knowledge of the translator and how relevant it is to the topic. When it comes to TM, the “knowledge” is the database that the translation memory is using.
So the secret lies in the process called “training” translation memory. This is when you lay the foundation of a good Translation Memory and afterwards a good quality translation.

And how do we know that? 1-StopAsia has been given the task to train translation memory for other companies and more important something knows as “corpora” or “corpus” – the source translations, which are used to build the rare Asian languages engines.

To make the final connection, Translation Memory is by far used in almost every translation company and by every translator. It is a way of improving productivity by removing the repetitive work from the translation process.

What is important to add is that in our experience the additional steps of editing and proofreading are still required in order to secure a good quality translation. That is why one of the most popular services nowadays is Machine Translation Post-Editing.

For more information on other topics related to our knowledge, you can read other articles from our BLOG.