Machine translation has come a long way over the past few years, and simultaneous translation technology is developing quickly, for both paid platforms and budget translation technology. Skype’s real-time voice translation, which can be used alongside a basic Skype subscription, recently added Japanese as its tenth language. Meanwhile, the Pilot earpiece, that will allow instant translations for conversations in real time, is set to be released later this year. On top of this, Google continues to add languages to its free translation service, and now supports a massive 119 languages.
However, there are still some aspects that human interpreters can do better than machines. In these instances, professionals believe that machine translation is just not good enough, or fast enough, to keep up with or take over human interpreters. Here’s why.
Humans learn more quickly than machines
As previously mentioned, machine translation works through deep learning to train artificial neural networks. This means that it can develop its language skills and improve as it listens to, and translates, more and more conversations. However, this process can be time-consuming, as the machine has to be able to learn the correct translations of phrases, especially when working out the context of a conversation.
Machine translation does not yet work in real time, currently operating with a delay of a few seconds. This is even true of the Pilot earpiece, which boasts the best available instant translation capacity for any device of its type. However, simultaneous translation can be completed with highly trained interpreters, a system which is usually used in high-level international meetings. Translators sit in a soundproof room to interpret in real time, while listeners hear the translations via headphones.
Interpretation is about more than just finding word-for-word translations
Earlier this year, Sejong Cyber University and the International Interpretation and Translation Association of Korea set up three machine translation programs against a group of human translators in a test to translate pieces of writing that had never been translated before. The test measured the accuracy, language expression, logic and organisation of the two translations, and found that the human translators scored higher overall.
While the machines managed to translate the sample pieces much faster than humans, the translated samples were less accurate. The VP of translation company Saltlux, Shin Seok-hwan, explained: “Though research shows machine translators can perform about 85 to 90 per cent of human experts, they still make absurd mistakes.”
While machines are starting to see the bigger picture and translate more complex phrases and longer sentences, they have yet to be able to take the whole document into consideration. Expert human interpreters, on the other hand, will have localised knowledge of the language, giving clearer and more accurate translations than a free or budget machine translation service.
Professional translators will have working and specialist knowledge of the industry
Human interpreters will have expert knowledge of the field in which they’re translating, and may often come from a trained background. For example, when dealing with legal or medical documents, the translator will come from the relevant specialist background, and be fluent in the two (or more) languages required. This provides a clearer and more accurate translation, and ensures that the finished document is error-free and, most importantly, makes sense.
Budget or free translation systems won’t have this (or any) specialist knowledge, which may lead to incorrect translations, as the AI system will need to have expert understanding of the industry before being able to recreate the language needed for the translated piece. Human translators can take everything into consideration during the translation process, making sure that everything makes sense, and that correct terminology has been used.
While machine translation has advanced and can provide clearly translated phrases in a number of different languages, it still lacks the specialist knowledge to be able to correctly translate whole documents without errors. The technology works by learning, meaning that over time, machine translations will become more and more accurate. However, human translators will always be able to translate more efficiently and accurately, without the need to “learn” new phrases.