Language JEM + The Machine

TRADOS 1997 SQ.jpg

Whilst having a new year clear-out, I found this old Version 2 TRADOS and MultiTerm installation disc from 1997! How far we have come in this field, but there is still some way to go.

I have been a fan of CAT (Computer Aided Translation) tools since my uni days. I had to teach my CAT tool the difference between oil used for fuel and oil used in food for a recipe translation into Russian. And indeed, I am still a huge proponent of Machine-Assisted Human Translation today.

Machines are here to help – be they translation environments, machine translation, QA tools, chat bots, speech recognition or AI. So in my view, a human will always be needed to produce accurate and high-quality content and translations.

People I know who use online translators often ask me why I bother to work as a translator when I have already been replaced by a machine. But be warned, these tools are not perfect yet. The mistakes lie not only in nuances, but some can also completely change the meaning. And if you don’t speak the source language, you may not even notice the error. Not to mention the confidentiality issue – where your content is stored on machine translation servers.

So, I am confident that I will not be replaced by a machine, but I do have tools that are my assistants. I am the one feeding and teaching them; I am the one in control. And I am the one responsible for the quality of the work I deliver – my reputation, and the success of my business, depends on it.


Here are two examples of critical errors I have come across recently:

1. A medical term looked up with an online machine translation page:

Czech source: antihypotenzní účinek

Machine translation into English: antihypertensive effect

Human translation into English: antihypotensive effect

In medicine, the difference between the prefix hypo (below, deficient) and hyper (excessive, above) is life-critical. And only using machine translation in a medical context could have the most serious of consequences.

2. A post translated on a social media site:

Original text written in English: "You're so unlucky"

Same text written in French (not translated) by the same bi-lingual person: «Tu n'as vraiment pas de chance»

French text automatically machine translated into English by the social media site: " you really don't have a chance "

So we can see that in any context the change from being “unlucky” to “not having a chance” is marked and significant.


There are many of these kinds of mistakes out there.  Of course, the machines are always learning, but for the time being this is still from human input.

If you have an important text that you need translated, please use a real person, otherwise you run the risk of critical mistakes and critical consequences.

If you are considering introducing tools into your workflows, then why not contact Language JEM to get some expert human input on your machine-assisted processes. We can help with: QA tools (Acrolinx), Machine Translation setup and optimisation, MTPE (machine-translation post editing), chat bot content, NLP (Natural Language Processing) and Corpus Linguistics. Get in touch.

Jemma Pullen