Machine Translation – Friend or Foe?
Machine translation – can it be a present or a future ally? Or is it and will it remain a formidable competitor for the translator’s profession?
Cheap, good and fast. Does machine translation solve the problem? Dare we hope that it can cover all three requirements?
Being passionate about project management, we consider machine translation post editing a new challenge. We don’t back off when it comes to testing out new ways to improve our work. Our aim is to always try out all kinds of programs, shortcuts, and anything that gives us more room to think. We can’t stand the saying “Don’t think, just do”, so whenever we have the opportunity to work less while being more efficient to have time for the fun part, the thinking one, we take it.
With regard to our project management activity in the field of translations, it comes naturally to find the most efficient methods to organize our projects. However, there are certain directions that we analyze at length and test them. Only after being certain that the one we opted for is the right direction or the best solution, we implement it. We didn’t dive headfirst into the vast area of machine translation (MT), but we didn’t avoid it either: we test different machine translation vendors with various CAT tools, different types of texts, we make sure that every tested solution is in line with the privacy regulations, we track terminological consistency, strengths and weaknesses; we keep in mind both the savings in time and the end result. We test and test and then we test some more. While navigating this endless world when it comes to translation automation (machine translation), we ended up being more and more curious about it, but also cautious before diving headfirst into this enticing universe – all of the above only to make the best long-term decisions.
Below you can find some elements that we considered since machine translation has increasingly caught the eye of the industry:
- Ethics of using MT – informed consent of the translator and the client
- Data privacy
- Which machine translation software should we use? Which is the best option for each project?
- MT review process – light and full machine translation post editing –best practices
- Best pricing method for machine translation (Machine Translation Post Editing – MTPE)?
- GPT-3 (Generative Pre-trained Transformer 3’) – how does it replicate the thought process of the human brain, and what is its impact on the translation industry?
- Pros and cons of MT
- Are we on our way to shift from computer-assisted human translation to human-assisted computer translation or will translator jobs simply disappear?
MT has been a very debated subject over the past few years because of the huge progress made in the field of automation and AI since it first appeared decades ago.
Neural Machine Translation (NMT) is the reason machine translation has reached a new level and gained ground in the translation field – with the help of artificial intelligence, machine-translated texts have become fluent, and programs have been modified in order to better adapt to them, so they no longer serve as simple dictionaries for words and expressions which ignore the context. There is certainly a long road ahead, but NMT seems to be a game changer for the translation industry, being preferred to alternative models such as “Statistical machine translation (SMT)” and “Rule-based machine translation (RBMT).”
Until recently, however, MT has only made notable progress for certain language combinations. This seems to change in the near future with the help of No Language Left Behind (NLLB) initiative, which aims to offer translations into and from around 200 languages, including rare languages such as Urdu, Luganda, and others.
We aim to clarify some of the issues around the subject of MT – the challenge is not to find answers, but to ask questions. To be open to new ideas, to take what is good and what suits us.
We hope you will join us along this ride!
