22 Jan AI TRANSLATION – WILL WE SOON BE REPLACED BY MACHINES?
Although it is commonly thought to be quite a ‘new’ phenomenon, machine translation has actually existed since the 1950s. Nowadays, it has improved dramatically, to the point that people are saying it is changing the translation industry, and some are even making the bold claim that it could replace human translators.
In this post we are going to look at how the different models of machine translation actually work, and by comparing its advantages and disadvantages to those of a human translator, discuss whether it ultimately does work and what its future could be.
How does it work?
Of course, the systems are extremely complex, and expert technicians and linguists use algorithms, encoding and decoding that would over-complicate this post if we were to expand on them in detail. Instead, we will briefly look at some simplified explanations of different AI Translation systems:
Initially, Rule-based machine translation (RBMT) referred to as the “Classical Approach” was used. This approach uses linguistic information taken fromdictionaries and grammarsof the source and target languages.
This was then replaced by Statistical machine translation (SMT) in the 1990s that works by analysing bilingual text corpora and data. It is a significant advancement from RBMT as the machine learns the linguistic rules itself, without the need for linguists to manually input them. It is considered the most traditional model, and is still widely used.
Neural machine translation (NMT) has departed from this traditional statistical approach, using artificial neural networks, its advantage being it uses single integrated models without the need for multiple. This technique is considered a breakthrough in AI translation, and is now being implemented by many platforms such as Google, Microsoft, Yandex and Facebook.
What are the Advantages and Disadvantages?
In terms of its disadvantages, apart from often producing translations that are incoherent, illogical or grammatically, semantically or lexically ‘odd’, there are many things that AI translation often does not consider or recognise; the use of words in context, wordplay and metaphor, the purpose of the text (marketing) etc. In the case of Facebook, the frequent use of slang and abbreviations challenges even the newest, sophisticated NMT. In other contexts, such as literary translation, AI translation is incapable of using initiative and creativity in the same way a human translator can. If a company requires a certain ‘style’, for example, this is not something that a machine would be able to provide.
That being said, not all AI generated translations turn out disastrous, and the advantagesof AI translation are easily identified – it is cheap as there is no need to employ human services. It is also automatic, meaning the translation is produced in real time. This can obviously be extremely advantageous to many businesses and people. On other note, it could even allow ancient languages of which native speakers no longer exist to be translated.
So does it work?
There is not a single answer to this question, because it completely depends on the purpose in mind. Whilst the translation quality will be expectedly lower, errors are generally overlooked as the main purpose for this type of translation is for the quick comprehension of a text. This is necessary in situations where time is limited and a company does not have time to wait for a human translation. It is also useful for the general public who may have no other access to a human translation and provides a quicker solution than consulting a dictionary. It may even be used to create a ‘first draft’ translation that is later edited by translators to ensure quality standards. This has proven to be a time-saving and cost-effective solution for certain text types and language directions, although it is in no way a replacement of what the human translator does.
AI translation can therefore facilitate and automate the practice of translation, however this in turn presents ethical issues associated with the role of the professional translator. With the emergence of these kinds of technology, issues arise of the image projected of what translation is; technology can stimulate misconceptions that the process is an automatic one and does not portray the complex semiotic, cultural and linguistic considerations involved in the process.
So what is the future for AI Translation?
Although there is a lot of talk that the development of neural networks for AI Translation is making major breakthroughs lately in many industries, its inherent limitations will never allow it to replace human translators. Perhaps we can expect more hybridity between machine and human translators in the future, as AI Translation continues to progress. However, perhaps companies will start to notice the difference in quality as AI Translation is used more widely, and begin to opt for high quality over low cost.
In conclusion, machine translation can be useful in certain situations, but it is important to be aware of what you’re using it for and the kind of result you will get. If you are looking for top-quality translations for your business, it is essential you employ professional translators to produce these for you. LexGo Translations is a translation agency employing professional, native translators you can trust to consider all the necessary contextual, linguistic and cultural aspects to produce flawless translations for any genre of text or language combination.