Generative AI has shown that it can keep up with – and sometimes surpass – established translation tools like Google Translate. Yet the technology still requires the input of professional human translators. In this article, we explain why (human) post-editing is still important for accurate translation, even in the age of generative AI.
Generative AI translation: Hype vs reality
With generative AI dominating the news for much of 2023, it is important to remain objective when it comes to the technology’s capabilities and to be acutely aware of its strengths and weaknesses.
As our tests have shown, the reality of generative AI translation is that it is similarly capable as existing AI translation tools like Google Translate. Sometimes it is more accurate, other times, less so – and it demonstrates different strengths and weaknesses.
In other words, it can’t reliably produce 100% accurate translation straight away and without the assistance and input of professional human translators – at least not for now. Language experts can use tools like ChatGPT to produce fast, automatic translations. However, the output stills need to be checked and reviewed before it is approved for use and publication. This process is called post-editing.
As the initial hype surrounding generative AI starts to calm down, the narrative will change and instead of being presented as a tool to completely replace human professionals, it will serve, as we have previously discussed in our blog, as a tool for helping to make language professionals’ jobs much easier.
Much like medical professionals are using machine learning to identify diseases faster and discover new treatments, professional translators will use generative AI tools to perform their jobs faster and more effectively. And this is something we have already introduced here at translate plus through our Generative AI+ package.
Why is (human) post-editing important?
Any translation project that uses machine translation tools requires a post-editing system to guarantee quality. The use of AI tools in language services like translation is nothing new; linguists have been using them for many years to speed up their workflows.
Despite all the advances in AI technology over the past decade, the accuracy of AI translation itself hasn’t improved all that much. While computational processing power has dramatically increased and innovations like Google Lens help algorithms translate text in images, the accuracy of translation remains relatively low.
This is why, in every project using a generative AI translation solution, a post-editing phase should always be added to the process – so that professional human translators assess the output and correct any mistakes.
Generative AI translation quality varies greatly, depending on the complexity of the source material and the languages involved. For example, translating simple text from English to Spanish should result in greater accuracy than translating from English to Mandarin Chinese (or Greek) – simply because English and Spanish have more linguistic similarities.
As a result, some projects require more time spent on post-editing than others.
Generative AI translation tools can help professional translators complete the early stages of a project faster. It can automatically translate a piece of text and allow the translator to jump straight to the editing stage, instead of manually translating everything from scratch. However, to ensure the right level of quality is achieved, the (human) post-editing process is as important as ever.
Do you have a robust (human) post-editing process?
If you’d like to discuss how the current breed of generative AI translation tools, with the help of post-editing, can help your projects, contact us today.