When we think of artificial intelligence, it causes many diverse reactions: fear, fascination, dislike, awe, uncertainty… Whatever our view on it, we have to admit that it is no longer just something of sci-fi films, it is very much part of our daily life. Take the case of Misty, the robot designed by the organisers of the Mobile World Congress as part of a pilot project launched in Barcelona, aimed at reducing the negative social effects of lockdown. Misty lives with the elderly Ana Margarita Fernández, greeting her when she wakes up, asks her if she has had her medicine and can even alert the health services if it sees something that it considers an emergency. Artificial intelligence could not be closer to home.

In previous blogs we addressed the matters of machine translation, the advance of technology and whether machines can replace humans in the field of translation. We rested comfortably on our conclusions that although machines showed impressive results in translating different types of text (certain types better than others), in certain aspects there was still much left to be desired and our jobs as translators were still secure. However, as we know, technology never sleeps, progress is continual and what machines and devices were unable to do yesterday, they are surprisingly able to do today…

Investigators and scientists have been developing what is known as “deep learning”. This technique seeks to give computers the ability to “learn”, by processing large amounts of data. Applied to translation, this technology looks for coincidences in its database and can produce a coherent and meaningful translation.

Metaphors and other hurdles

However, the problem still lies in the difficulty in giving AI something inherent to the human being: common sense. Although a computer database may be extensive, as human beings, we are a walking, living database, enriched by all the life experiences we have acquired, together with this inert common sense which helps us to distinguish, to sense and to choose, and make intelligent linguistic decisions.

A high-quality translation is not just a matter of syntactic or grammatical correctness. Language involves contextual information that goes beyond the mere words themselves. There are many aspects of language that a machine cannot detect, such as metaphors, humour and irony.

Let us consider one of these items. A metaphor is a figure of speech that draws a comparison between two apparently unrelated things. It describes a person, an object or an action in a way that is not literally true, but it helps to convey an idea or make a comparison, adding depth and colour. Did you spot that we used a metaphor earlier in this article? In the second paragraph, we said that technology “never sleeps”. Here technology is given a human quality, we know that it does not literally sleep but the phrase transmits the idea that it is in constant progress and development. Care needs to be taken with metaphors, since although sometimes there is a similar metaphor in the target language and can be easily transferred, they are often tied to source culture and shaped by socio-cultural beliefs and attitudes. If there is not an equivalent target language metaphor, another method must be sought, such as using a simile or opting to translate the sense, although this may lead to a loss of the emotive force that was achieved through the metaphor. Metaphors may involve imagery and references that may be confusing or even offensive if transferred to the target language.

In a nutshell…

As well as having trouble recognising linguistic aspects such as these, and lacking the sensitivity to deal with them, AI cannot spot issues that may be rare or unique, that are not included in its data. We can safely say that AI is not able to entirely automate the translation process and produce a perfect translation. It certainly has much to offer in speeding up processes and offering a good quality pre-translation when dealing with large volumes and high demand, however AI and machine translation should still be a complement and an aid to human translators.

Misty may be effective in making sure Ana Margarita takes her medicine, but nothing can replace a warm smile and hug from another person. In the same way, language will always be communication between people and cultures, and translation will always need that human touch.

Header image by Hello I’m Nik on Unsplash