What is translation today?
Updated: Sep 17
“What is this translation? Google Translate?” The Love-Hate relationship people have with Machine Translation engines is well anchored in mentalities. While it certainly developed on fertile ground, Machine Translation is not what you think it is. Not anymore. Oh! And where is the CAT gone?
Computer assisted tools vs Machine translation
It would be pushing on open doors, writing that the face of translation has drastically changed over the past few decades. However, what exactly has changed? When you ask people who are not working in or with the industry, not many have a clear idea. Most of them will mention machine translation and will utter a vague premonition that in the close future, translation will be performed by computers and not humans.
So how translation is done today? Are translators an endangered species?
The big change, that has already happened is called computer assisted tools (or computer-aided) known also as CAT tools. CAT tools appeared in the 80s, and progressively their use has been adopted by most of translators and I will venture to say by all translation agencies. There is a great number of CAT tools out there, with different interfaces but very similar features.
CAT tools are time savers and they are also useful tools for the maintenance of terminological consistency throughout projects, which contributes, in the long run, to high quality results.
How do CAT tools work? Are they a threat to human translation?
No, humans are safe with CATs as the outcome is a 100% human translation. A CAT tool, in a nutshell, is the equivalent for the translator of the IDE for the software developer, namely, a work environment in which translators have handily access to all their resources: source text, glossary and the unmistakable Translation Memory.
The Translation Memory is a database of translated sentences. Every sentence translated in the project is stored in the memory. Every time the translator starts translating a sentence, the CAT tool searches the Translation Memory for a similar sentence that would have been translated in the past, and proposes to the translator all the results, source and target, for them to judge if an older translation can be reused.
This way, the translator has a quick and easy access to already translated segments that they can accept as they are or adapt according to the context. Furthermore, the translator can attach to their project a Termbase (glossary) containing terminology approved by the client. In a similar way, each time a term that already exists in the Termbase appears in the source, the term is displayed in a side window and the translator can insert it in their translation (or not).
The benefits of this system include processing larger volumes in shorter timeframes, with increased consistency and at lower costs. The time argument is obvious as the translator does not need to type, let alone translate, repeated segments found in the Translation Memory. Likewise, the consistency stems from the fact that by having easy access to everything that was previously translated, the translator can remain consistent with previous works and within the current project. Lastly, as the cost of segments already present in the Translation Memory is lower than that of new segments, there can be a very positive impact on pricing. CAT tools are helpful for translators, clients and agencies. The main criticism encountered among some translators is that CAT tools unfairly pulled their prices down because of this split in rates between new words and already translated segments (called fuzzy or 100% matches).
Although this process is not very widely known of the general public nor translation buyers, it has been a standard since the late 90s. Recent developments have added more functionalities, in particular, connectivity with Machine Translation engines.
The next revolution that is happening now is definitely Machine Translation (or Automated Translation). The concept has been around for decades, back to Cold War in the 60s. Long ridiculed in reference to its very unlikely outputs, Machine Translation has, in recent years, thanks to technological progress, achieved quite promising results, especially lately with the development of Neural Machine Translation. These engines are designed to mimic the human mind (as opposed to its predecessors the Statistics and Rule based engines).
It is important to remember: forget anything you knew about Google Translate. Properly trained Machine Translation engines can produce very satisfactory raw output.
Hence, the worry that one day, humans will be replaced. However, depending on the type of projects, past the fully automated production of the translation, a second phase of human editing can be necessary. And a new profession is born: the post-editor.
Will post-editors replace translators?
Here again, it seems translators are not doomed to all convert to post-editing, at least not in the near future. Although, the use of Machine Translation is in fantastic expansion , full human processes are still needed wherever a trained engine is not available, or MT is simply not trusted.
Look out for our next article about big changes in the translation world. Next, we’ll talk about Translation agencies.