Do human translators have a future?

Updated: Dec 1, 2020

Can we rely on machines to translate our most complex communications? Will human translators one day be out of a job?

All around us, industries are becoming rapidly more automated and being constantly refined for greater efficiency and the field of translation is no less affected by such changes. Google Translate is just one automated translator that has undergone significant development over recent years and is, perhaps most compellingly, free for public use. Indeed, almost everyone we know with smartphone or computer access has turned to Google at one time or another to translate a simple segment of text. We might reasonably ask, 'Is it really necessary to pay a translator? Can an automated translator do the job for me?'

These are common questions. If we already have automated means of translating and if computers are improving all the time, why pay the significant fee for a person to do the work?

Machine translators can certainly be useful and effective in some situations. In this article, we will see what sets a human translator apart from an automated translation tool, why human translators may be considered irreplaceable and how automated tools support the work of human translators.

What happens in translation?

The existence of dictionaries is evidence that every word has at least one definition or explanation as to its use. Multilingual dictionaries, which contain equivalent words and terms in different languages, might seem to make the case that translating from one language to another is simply a matter of choosing the equivalent word. Indeed, there is a commonly held belief, although vague, that translation is a mechanical process of stringing together equivalent units, perhaps even in the same order that they appear in the original language.

What dictionaries do not show is that language is far more complex than stringing together simply-defined words. Let's understand a few of the challenges involved in the work of translation.

Complex definitions

For one thing, words are not always easy to pin down to a single, clear definition. A word may have multiple definitions.

Let's take the example of the word 'call' as in the sentence, 'I will give you a call tomorrow.'

'Call' may be used as a noun or a verb. It may relate to any of the following concepts:

- shouting

- using a phone

- making a prediction

- visiting as a medical professional

- stopping by for a quick visit

- referring to an existing name or bestowing a name on a person, place, object or idea.

These are only a few of the definitions associated with a rather simple word.

Context is everything in understanding many of the words we use. Even in a clear context, there may still be ambiguities. Quality translation requires effective judgement as to what is the most likely meaning of a given word in context as well as the most accurate equivalent in the target language.

Machine translation relies on algorithms, which can allow for a degree of effective context analysis. However, machines cannot offer intuitive reading for some of the more subtle ambiguities or even for some of the large and obtrusive roadblocks.

Unique words and phrases

Every language has its own means of expressing aspects of the human experience and the world we live in. Some languages encode specific ideas that are not even named in other languages. One example is the German word abendrot, which loosely means 'sunset glow' or the colours of a sunset. Although English speakers have seen sunsets and understand this concept, we don't have a single word for the associated colours. Instead, we need to arrange words to describe the same idea or experience, such as by creating the simple phrase 'sunset colours'. Machines can only go so far in tackling terms with no simple equivalent in the dictionary, while the human mind has the creativity and imagination to circumvent these problems.

What about ideas that simply do not exist in other languages? Many, even most, English speakers will be unfamiliar with the word brigadeiro. On asking for the meaning, we will learn that it is the name of a common Brazilian dessert made of condensed milk combined with other flavours. A machine translator faced with this word will only be able to follow the specific instructions it is given. If no English equivalent has been input in the system, the machine will probably give the Portuguese name. This leaves the reader to do their own research and hope they can find an explanation for the term outside of the translated text. On the other hand, a human translator is able to assess the situation and will know whether it is appropriate to describe the dessert, replace it with a culturally equivalent dish in English, or simply give the original name.

Idiomatic expressions can be even more complicated. These usually describe universal experiences but sometimes in fascinatingly unique ways. Some idioms have well-known equivalents that can be programmed into machines as single units. For example, raining cats and dogs in English is often equated to raining cords, (il pleut des cordes), in French. If this particular expression has not been fed into the machine in advance, the machine will do its best by creating a literal, word-for-word translation. The result will either be unintelligible to the reader or, at best, require an open and creative mind for a speculative interpretation.

At least in this case, French had an equivalent term for raining cats and dogs which could be programmed in advance. What if someone told you, in translation from Afrikaans, that the jackal was marrying the wolf's wife? This incredible idiom describes when it is raining and the sun is shining at the same time. We don't have a similar idiom for this situation in English, and this presents a challenge to the translator. We could describe the situation in plain terms, which loses the fun metaphor of the original expression. Or, we could translate the expression literally into English with context or an explanation to give it the desired meaning. We could even create a new expression with culturally familiar terms if wolves and jackals are uncommon in the target location. Each one of these solutions requires careful consideration beyond what a machine can offer.

Different language structures

We have seen that not all languages have equivalent words and phrases. If we 'zoom out' to the bigger picture, we can see that languages also have different ways of organising their words and phrases, which needs to be taken into account in translation.

A basic part of sentence structure is where we place subjects and verbs. You may wish to click here to learn more about subjects and verbs and how they can be ordered.

In English sentences, we place the subject before the verb, like in the sentence 'she drives a car'.

In Spanish, this structure is more flexible. The subject might move around the sentence or even be considered unnecessary, leaving a phrase that could be literally translated into English as simply 'drives a car'.

Machine translating the sentence from English into Spanish will probably mean keeping the subject, even if the resulting sentence seems unnatural in Spanish. On the other hand, translating the Spanish sentence drives a car into English leaves a gap where the machine must guess at how to fill in the missing subject. If the phrase is isolated with no context to guide the decision, the machine is likely to choose the subject he by default instead of the intended she. These mistakes may occur with machine translation even when there is surrounding context due to the limitations on how machines can interpret nuance and implied meaning.

What about existing English expressions that don't even adhere to the standard rules of English grammar? Phrases like 'long time no see' and 'no can do' are examples of language that purposefully flouts the rules. English speakers know what is meant, but an automated translator would need to be kept updated on language change and pop-cultural references to have any chance at all with these irregular structures.

Cultural perspectives

We have noted that each language has its own way of naming and describing common objects and ideas and that difficulties arise when these concepts don't exist in another culture. Here is another problem to consider. What about concepts that exist and are named in both languages, but are understood very differently according to cultural perspectives?

One example of this might be to say that someone is independent. In Western cultures, personal independence is generally valued. It defines a person who is strong, motivated, and unlikely to drain resources from those they are close to. In this context, independence is a positive trait.

This is not always the case around the world, however. As Eastern cultures tend to be more focused on groups and shared experience, independence can look like selfishness and rejection of support networks.

A dictionary may determine a simple equivalency of meaning, but probably will not be able to capture the value assessment afforded by cultural usage. A machine translator may awkwardly present a term with a different connotation, thereby distorting the intended meaning. A human translator will be able to weigh up the values at stake and place the text in its wider context. For example, in a eulogy, would it be appropriate to describe a person as strong and capable? Of course. What about remembering their selfishness? Not so much. The human translator knows when and how to integrate information in its greater context.

Translating between the lines

We have looked at words, phrases, idioms, structures and cultural values. It would seem that the case has been won in favour of the human translator.

While machine translation ultimately functions on the basis of transferable units, human communication and experience are overall deeper and more complex. Machines don't have a sense of humour like we do. They don't understand tone or allusion. This failure is made more complex by the fact that every culture has its own codes of nuance and sense of what is funny and what is not. Jokes, poetry and famous quotes all need to be translated in a way that gives the same effect in both languages, whatever that may be. Is the poem intended to be as beautiful in its translated form as it is in the original? The translator may choose to make stylistic choices in recreating a poem to be beautiful in its own right. Is the poem to be translated for its historical merit, incorporating all references to people, places and objects as factually as possible? The translator may then sacrifice rhymes for the sake of accurate representation of names and information.

Do machines have a place in translation?

Machine translation cannot do everything that a human translator can. However, the art of translation is not, as a whole, immune to the impact of technological advances. In fact, automation has already moved the industry forward significantly in a supporting role to the work of the human translator.

One example is common but unreliable. We have seen that automated processes can return a basic, unpolished translation at varying degrees of acceptability. Why not then run any text first through an automated tool and give the translator the easy job of simply 'tidying up' the translation? Perhaps unsurprisingly in our world of speed and efficiency, this is a common solution that is popular among clients wishing to pay the translator a lower rate for their part of the project. The translator's work of looking over a prior machine translation is called 'post-editing'.

Unfortunately, this machine-led shortcut is generally not as effective as it may seem. Even if the text has not been left an incomprehensible mess from the automated translation, a conscientious translator would still do well to check each word carefully against the original in case of mistranslations. An informed client will know the boundaries of reason in making such a request and how much the work is worth. Otherwise, and as is often the case with such projects, the translator is likely either to work beyond the terms of their payment or return a shoddy translation.

Automation is not all about cheapening the industry. In the more exclusive realm of specialised software, translators are using technology like never before to manage their work and boost their efficiency while never relinquishing their role as the brains behind the work. Computer-Assisted Translation programs, commonly known as CAT tools, have transformed the work of modern translators. This computer-run software is used to extract and segment text for easier processing, analyse text, remember and reuse previously approved translations, streamline tasks and process the text into the desired format and file type. Translators pay big money for subscriptions to and training in the best quality software, which is essential for any who want to be taken seriously and meet the high expectations of informed clients. The CAT provider currently leading in pricing, popularity and the cutting edge is the acclaimed SDL Trados, which is a household name within the industry.

Translation, like every field, is growing and changing as a result of new technology. However, the changes in this area are not centred on phasing out human input. The discerning human mind is still the most competent translator available and is likely to remain so indefinitely. The work of translation is a complex task and translators need to balance a lot of factors effectively to create a quality text. Machine translators are limited in their resources and their ability to assess the priorities of a given translation task. Nevertheless, machines can help make our work more streamlined and efficient, and they already have an important place in the industry's development. Now, go ahead and give your (human) translator a raise. They've earned it.

Do you have a question about linguistics or the language industry? Ask Shana to learn more.

24 views0 comments

Recent Posts

See All