Machine translation systems have advanced rapidly in recent years and have become sophisticated technologies that allow excellent results to be obtained without the need for extensive proofreading. However, in certain contexts human editing and proofreading is fundamental to ensure translation is of the highest possible quality. In this article we take a closer look at this issue to discover the advantages and disadvantages of both types of translation and how they complement each other.
Machine translation vs. human translation
A machine opens a document and automatically starts translating it without stopping. It does this in a highly efficient way, based on what it has been learning and analyzing 24/7. It goes without saying that Machine Translation (MT) translates words with a high level of accuracy and at an extreme speed – something that can’t really be achieved at the human translation (HT) level, as a person usually translates around 2,000/2,500 words in a day.
But humans add value when it comes to idioms and expressions, i.e ways of speaking and constructing complete sentences. A translator brings their personal touch to each paragraph. In addition, the translator is familiar with the semantics and terminology of the language. As expert developers of translation engines, we know that machines are able to specialize quickly with what is known as parallel data, then a human comes in to deliver context.
For a translation to be as good or better than an original version, the context of the original text or audio must be fully understood, and this is where human translation plays a very important role. Perhaps then, we could say that the role of the translator is morphing to become more of a "quality reviewer" of large volumes of machine translation as we will see below.
When is a human translator needed?
As technology advances, neural machine translation systems are getting better at processing language. Based on tools such as machine learning (ML) and deep learning (DL), they provide machines with the ability to learn from parallel data, without being explicitly programmed. As they learn, their ability to express themselves improves, as does their ability to translate.
Certainly, it’s the expert hands of programmers and artificial intelligence experts who administer, manage, select and prioritize data and how algorithms look at certain parts of a sentence, or tell them what vocabulary or take into account but its the work of the human translator as the linguistic expert who provides situational knowledge, who resolves polysemias and contexts. So translators are needed when thinking about quality in cases where:
- We want to avoid or eliminate translation errors.
- We understand that the context will influence the translation.
- The translation is in a specialized field.
- The text will be published and the accuracy of the translation is important.
According to experts, the last two points depend on the amount of available data from the specialized field that has been fed to the translation engine. So if we had a relevant amount of parallel data from medicine, agriculture or the legal sector, our translation engine will specialize in them having seen relevant examples.
A professional translator avoids translation errors, or errors of polysemy, and by the same token corrects these errors in the source text. In addition, the translator provides text adequacy to the field's own style.
With the imminent digital transformation that companies are undergoing, along with the growing demand for translated content, these need to be increasingly specialized yet varied enough to have an impact on those who read them. This is why more and more companies are opting for automated translation of only the most general parts of their content, while the more personal and/or specific parts are entrusted to humans.
Recommended reading: Post editing of machine translation
The advantages of machine translation with post-editing
There is also a hybrid form of translation. A process by which human and machine both play a role. In other words, a human post-editor improves the quality of automated translation. In many cases, post-editing is a very good translation option. Here, the post-editor takes advantage of the speed and availability provided by the MT, and then turns their attention to reviewing and post-editing the translation.
The more specialized the text, the more the need for a human translator to scrutinize the document. The weight of the human post editing also sometimes depends on the quality of the MT and its output. It may be more or less necessary for the translator to post-edit, sometimes even having to directly translate the text again.
Translation with post-editing has many advantages:
It is very cost-effective if the translated text is understood without much difficulty and is accurate, in the absence of minor details, it would not be worthwhile totranslate it again
It unifies the concepts and solves those mistakes made by the machine, adapting those expressions that are not literal and verifying the accuracy of the way in which the information transmitted by the MT
Both technologies complement each other and offer a balance between efficiency and translation quality, being a much faster option than human translation in most cases
Most professional translators work faster as editors than as translating from scratch.
By saving human translators hours of work, it's actually one of the fastest and most cost-effective ways to achieve a result that is up to par.
Pangeanic: human and machine translations adapted for you
Any company with an international presence needs top-notch translation services for its written translation services. In order to communicate effectively in the global marketplace, accurate translations are required across all physical and digital channels.
In a world that continuously produces more and more content, how do you translate efficiently and effectively? As machine translation continues to evolve, Pangeanic has taken advantage of the latest digital technologies and assembled the best certified talent to bring you the high-speed, cost-effective translations you’re looking for.
Pangeanic's Deep Adaptive technology allows engines to be trained on a one-to-one basis and with near-human parity. This allows us to offer the best MT, anonymization and summarization services, among others.
Pangeanic provides you with the translation option that best suits your needs, from pure machine translation to post-editing services (MT and human proofreading).
Depending on the customer's needs and the quality of the machine translation engine
Pangeanic adapts its intelligent translation system by finding the perfect
formula combining both artificial intelligence and natural language processing.