Since its introduction in the 1950s, the use of machine translation has been increasing and it is one of the most discussed topics in the translation industry. Machine translation (MT) refers to the automated translation of a text from one language into another.
During the age of social media, the amount of content created daily has grown enormously, and it can be difficult and expensive for translators to process such large volumes. That is why, in a society that demands more content more quickly, MT can be a solution that allows companies to translate content, which would otherwise not be translated and which would remain inaccessible to certain users, into more languages in a more cost-effective way, as MT can process millions of words in a relatively short period of time.
The evolution of machine translation
MT started as Rules-Based Machine Translation (RBMT), in which linguistic rules and lexicons were encoded by humans, then evolved into Statistical Machine Translation (SMT), which leveraged machine learning in order to extract rules from bilingual input and select words that are likely to appear together in a translation based on statistical models, which delivered better results than RBMT. The newest approach, Neural Machine Translation (NMT), which uses deep learning to extract information about the way words associate with each other and modifies itself to increase the likelihood of correct translations, has shown great improvements in fluency and accuracy.
However, despite all of this progress, MT still has its limitations and the results are far from perfect; for example, MT might leave words untranslated, add or omit words, use the wrong word order or punctuation, and deliver mistranslations. Even today, MT cannot match the quality of human translation. High-quality translation requires knowledge of the subject matter, understanding of the context, cultural references, nuances, and tone, for instance, which machines are not able to recognise. This is why most MT output requires post-editing.
What is post-editing
Post-editing refers to the human editing of the raw output generated by machine translation systems in order to achieve the desired final result. The quality of MT greatly depends on its correct application. In order to get useful output, machine translation engines must be trained. While there are generic engines, they can also be customised for specific projects to adjust to your own terminology and style. The extent of post-editing will vary from project to project. Depending on the customer’s needs and the intended use of the content, different degrees of post-editing can be applied.
Types of post-editing
For content with a short lifespan, low-visibility or low-risk, such as FAQs, comments under a blog post, or internal communication within a company, light post-editing or post-editing to understandable quality can be the right option. In this case, only major mistakes that affect the comprehension of the text, such as mistranslations, omissions, and additions are corrected, while certain grammar, spelling, and punctuation mistakes, for instance, are left untouched, unless they affect the meaning, and no corrections of stylistic nature are made.
For content intended for publication, full post-editing or post-editing to publishable quality is the way to go. This kind of post-editing follows the same principles as human translation. Grammar, spelling, and punctuation errors are corrected, and style, fluency, and terminological consistency are improved. The final result of this type of post-editing should read just as well and naturally as human translation, and it requires specialised translators in the specific domain.
Is it always a good idea to use MT?
Given that the purpose of MT is to save time and effort, it is important to mention that not all content is suitable for MT, and there are cases in which its use is not recommended. For example, the translation of content that highly depends on nuances or cultural references, such as marketing materials or literary works, might be better done with traditional human translation.
One domain in which MT can be quite useful, however, is technical translations, where there is far less nuance, and words might have fewer connotations. While this type of translation still comes with the complexity of the respective fields, it is more straightforward, and the more specific the translation of a word is, the more likely it is that the machine will choose the right translation.
Overall, it is important to remember that the goal of MT is not to replace human translators, but to increase productivity. While there is no denying that this technology can be useful, post-editing is something that needs to be done. It will help you achieve the high level of quality you require for your content and give your customers the professional impression for which you are striving.
Do you already use this technology? Are you looking for a Spanish post-editor? Please feel free to contact me and I would be more than happy to assist you.
Image source: t20-P3AYKR|© tampatra | twenty20.com