As a translation student, I have always been intrigued and interested in CAT tools. In this first year of my master’s degree, I have been lucky enough to have classes to learn how to use CAT software such as SDL Trados. I also learned how professional translators use them, because it’s an indispensable tool in our profession, no professional translator does his or her job without using one of these softwares.
How do CAT Tools work?
When you are starting as a translator, or when you’re translating a document in a field you are unfamiliar with, you start from scratch: no termbase and no translation memory either. Therefore, the CAT tool cannot give you a segment called “fuzzy” or “perfect match.” It will automatically be a “no match.” Thus, you have to create them yourself, a first argument that makes us believe CAT tools will never surpass human translators: the machine cannot recognize terms and add them to the database automatically, which means translators have to do it as soon as they see one term that does not appear in the termbase they themselves created, and they have to do it during their whole career. As for translation memory, it’s a little better (at least for non-beginner translators): when you save a target file, the software remembers it and can retrieve similar or identical segments for future translations. There are a few factors to consider, but we won’t go into the details.
The second argument is the “align document” function: the software performs a kind of “pre-alignment,” but the result is sometimes far from being satisfactory, and the human translator must do it as well after the machine and confirm, connect, disconnect, confirm, or reject (among other things) the alignment. It is also possible to integrate a machine translation system into a CAT tool, but this is unreliable, and human translation is better.
CAT tools require human intervention, whether for pre-editing or alignment, as mentioned above, or for post-editing.
These are some of the features that the tool cannot do well on its own, but there are many more. If you want to know more about CAT tools, you can read the article, “What role will CAT tools play in the future?” written by Ophélie Tournier.
Artificial Intelligence in Machine Translation and CAT Tools
In CAT tools, there is also Artificial Intelligence. The latter was first integrated into machine translation; therefore, we will start by seeing how AI works with Machine translation, and then how it works with CAT tools.
Machine translation is possible thanks to Artificial Intelligence, which is getting better and better with each use, because it collects and analyzes data every time a translation is done with a machine translator. A lot of progress has been made since the beginning of the use of machine translators, but it is not going as fast as the experts would like.
Nowadays, the type of machine translation that is used is called Neural Translation (because it works like human neurons). Part of Artificial Intelligence is what is called “deep learning,” but what is it exactly? It’s a deep method stemming from the concept of “machine learning” which is specifically based on the notion of artificial neural network. An artificial neural network has a non-linear structure and looks like a network of sets of information execution units (which represent neurons). These units are layered and linked together via connectors (called synapses). That is when this information is processed through propagation patterns of activation of these units, which activates beyond a certain threshold. “Deep learning” makes Artificial Intelligence autonomous because it allows it to integrate new rules on its own and thus create ever denser and more complex layers of neurons.
The problem is that this is very expensive, and since Artificial Intelligence is used in many different sectors, deep learning or machine translation are not necessarily the main fields of researchers.
If you want to know more about machine translation and Neuronal Translation, you can read the article, “Why Is Machine Translation the best upgrade CAT Tools can hope for?” written by Luçanne Raharinosy.
In SDL Trados there is an option called “pre-translation” which allows the software’s Artificial Intelligence to offer the translator a post-translation and therefore, save a lot of time. This way, the translator does not have to write all the sentences, and only has to correct the translations proposed by the Artificial Intelligence.
Translators started using CAT tools such as SDL Trados in the early 2000s; a lot of progress has been made in 20 years, but there is still a lot to do,
Because CAT Tools can’t work on their own and translators need to make changes or add things (e.g., translation memory and termbases), and a lot of things have to improve in the field of Artificial Intelligence area, CAT Tools will not replace human translators anytime soon.
Chrystelle Bongrand – M1 TSM (2021-2022)
La Traduction de A à Z 