In the translation industry over the last few years, there has been a lot of discussion about machine translation, and often namely Google Translate. Its progress is being watched carefully by translators across the world, whose whole approach to work may have to change the more success this technology achieves. Machine Translation or MT may be the translators’ foe, competing for their jobs and undermining the quality work that is produced, however currently it is still not a tool that can be employed for high-quality, accurate translation. Computer-Assisted Translation (CAT) on the other hand is the translator’s friend, helping them to become more efficient at their work. It is often confused with machine translation, when actually its technology and success in application is very different.
This week, eSense Translations’ blog looks more closely at CAT and how it is an effective tool for translators, compared to their enemy of MT.
A bit of history…
Computer-Aided Translation (CAT) tools have actually been around for quite some time. It was in 1950’s and 60’s when the question was first posed whether the previous work of translators could be stored, organised and then re-used to assist with future documents.
In the 1980s, Trados were the first to develop a product with this functionality and have been a strong leader in the market ever since. Recently, other competitors, including Google, have created similar technology. Nonetheless Trados, now bought out by SDL, remains the go-to company not only for many translators, but also for language service providers and even corporate departments themselves.
CAT is a fantastic tool for translators, increasing their speed and efficiency, as well as their consistency and quality of work. Below, eSense Translations looks at some of the features that are included in the SDL Trados products, but similar tools will appear across most CAT software.
This is key function of CAT tools. Translators (or larger organisations) are able to record and store their translated content for future use. The database will generate stored ‘matches’ on new content that is being worked on to ensure consistency across all of a translator’s work. This becomes particularly helpful when text is technical and specific translations of certain vocabulary are required for accuracy. As the database in the translation memory grows, the faster and more efficient a translator’s work becomes. And as the translator still has the input into the translation, the quality of their work is not put at risk.
This feature provides translators with ‘intelligent suggestions’ as you type. It monitors what is being typed and after a few characters provides suggestions, based on the context that the word appears in. The translator then has the option to select the desired word from the list, saving them typing time. The time saved may seem small, but over the translation of a large document, these few seconds saved start to mount up and become important. I would compare this feature to be a little like predictive text, but hopefully more useful and ‘intelligent’ than the random words that can sometimes be generated here!
This function allows you to search the translation memory for a particular word, sequence of words or phrase. These translation ‘units’ will appear in a search window, displaying how they have been previously translated. The concordance search can be completed on both source and target text and assists with ensuring consistency throughout translations.
The translation memory assists with the storing and reusing of segments of text, however often translation is not that straight forward and text from one source to another will not match exactly. Fragment Recall allows for the matching of fragments at a sub-segment level. This helps greatly for instances like these when translators encounter ‘fuzzy matches’ or even a fragment defined as a ‘no match’.
This tool allows translators to even use their previous translations, which were not created using CAT, to build their Translation Memory. This means that any of the translator’s work can be used to assist with future translations, removing the need to create the database from scratch.
SDL Trados has a function in its software that is able to identify the ‘fuzzy matches’ and then repair them, entering the best match possible into a translation. This tool is known as Fuzzy Repair. This removes the need for the translator to do a post-edit and make the changes themselves, again claiming to increase speed and efficiency and also reducing tedium. The Fuzzy Repair function uses intelligent sources to make these judgements, such as machine translation and translation memories. However, is this stepping into the territory of machine translation? The important differential of CAT tools over MT is that the human translator retains control; they effectively sign-off the finished translation as being correct. If translators are using the Fuzzy Repair tool to post-edit, relying on it to be accurate and then sending the file out without any further checks, then we are losing that human intelligence that is keeping straight machine translation from being a total success.
Comparison to MT
Machine translation is very different to translators using technology, such as computer-assisted translation to improve their working processes. The aim of MT is to have limited to no input from human translators. The problems in the past with MT has been that the resulting translations have been very literal and therefore documents are produced that do not read well and often make little sense.
There have been improvements made recently using a system called neural machine translation. This system uses large artificial neural networks to try and predict sequences of words, instead of translating word for word. The result is better, but still far from perfect!
For a further insight into the evolution of machine translation, eSense Translations has written previously on the MT race and how it could affect the translation industry.
CAT is a very useful tool for translators. It assists with their work, allowing them to work more efficiently and consistently. The technology enables them to take on more work and also offer their clients better value. MT, as it stands, is still aiming to squeeze out translators by providing a cheaper service, but at a far lower quality of work.
By Lorna Paice