top of page
  • Writer's pictureZoe Womack

Machine Translation & AI: Are Human Translators Becoming Obsolete?


A human hand and robot hand typing on the same laptop simultaneously.

In recent years, there has been noticeable advancement in the capabilities of machine translation (MT) and artificial intelligence (AI), sparking considerable debate about the future of the translation profession. There is growing concern that these tools may render translators obsolete, and that the amount of work we are offered will gradually decrease.


For instance, about a month ago, I came across this post on the popular online forum Reddit, where a user posted about their concerns about the future of the profession in the subreddit r/translationstudies.


A post on Reddit by a user called AlbertSnap. It's titled "Should I continue pursuing a career in translation?" At the time the screenshot was taken, it had 26 comments. The full text can be accessed via the link above.
A screenshot of a post on Reddit expressing concern about the future of the translation

While some users encouraged the poster to quit so there would be "less competition", most of the responses were optimistic. Many pointed out that the role of the translator is not becoming redundant, but merely changing.


Others pointed out the limitations of MT and AI, and urged the poster to consider specialising in niche areas that machines are either:

  1. Not trusted to translate due to the important nature of the texts (eg. medical and legal translations); or

  2. Unable to translate well due to the creativity and localisation required (eg. marketing and literary translations).


I also added my thoughts to the comment section. You can read it on the Reddit post here, or in the screenshot below:

My comment on the Reddit post mentioned above. My Reddit username is ezotranslation, and the comment had 33 upvotes at the time the screenshot was taken.
A screenshot of my comment on the Reddit post.

You can also read more about my thoughts on the unreliability of machine translation on my website here.



Of course, MT and AI have their advantages as well as their disadvantages, which I will discuss below.


Advantages of Machine Translation & AI

  • Speed Automation means that machines are capable of processing large amounts of text in a matter of seconds. Much faster than a human ever could on their own.

  • Efficiency Machines are capable of translating texts into multiple different languages simultaneously, something that is impossible for a human translator.

  • Cost-effective & easy to use Machine translation services such as Google Translate and DeepL, and AI such as ChatGPT are free to use and are fairly intuitive to use.

  • Gives you the gist While they are not reliable for producing publish-ready content, they can still be useful for getting the gist of a text.

  • They can learn Machine translation and AI are always learning and making improvements based on the input they receive from their users.

Disadvantages of Machine Translation & AI

Despite their advancements in recent years, MT and AI are still a long way away from matching the quality of a professional human translator.


  • Literal, word-for-word translations Machine translation is only able to produce literal, word-for-word translations. This is usually highly inappropriate, especially for creative and customer-facing texts such as website content and product information, where businesses want to evoke a certain response from their target audience. Since machines are unable to take into account the text type, translation purpose, and target audience, they cannot adjust the translation or tailor it for a client's specific needs in the same way a human translator can.

  • Errors MT and translations produced by AI frequently include mistranslations, omissions, repetitions, and even sometimes unwanted additions to the text! This is especially true when translating between languages that are very grammatically and culturally different, such as Japanese and English. While a human translator can "read between the lines" and find context clues to adjust the text to be appropriate, machines will often try to force a literal translation, often resulting in mistranslations and unnatural language in the translation. Here is an example of Google Translate attempting to translate a section from this Japanese article about the shortage of fries at fast food restaurants during the pandemic.

An example of Google Translate incorrectly rendering Japanese into English. The English reads: "'Please have a hamburger and a set of French fries.'  'French fries' that I usually eat casually. However, the restaurant industry, which provides the products, is currently suffering from a shortage of raw material inventory, and some stores are unable to sell some products. What the hell is going on with Potato? How long will this state last?"
An example of Google Translate incorrectly rendering a Japanese text into English.

For those interested, the first sentence is actually a customer asking for a hamburger and fries in a meal.

In Japanese, subjects can be omitted from the sentence if it's considered to be obvious from the context. A human translator can use their language and culture expertise, as well as context clues in the text to determine the intended meaning and find a suitable subject to insert into the translation. However, machine translation, such as Google Translate, is unable to do this, resulting in translations full of mistranslations and grammatical errors, like in the above example.

  • Struggles with certain terms Machines struggle to translate things such as culture-specific terms, borrowed terms, idioms and nuances in texts. This is something that can only be done by a human translator who is fluent in both the source and target languages, and is familiar with both cultures, preferably having spent years living in the country where their source language is spoken.

  • Unreliable and damaging Important texts such as medical, financial or legal texts, cannot be entrusted to machine translation or AI for translation, as any errors can cause major issues. As Frédéric Ibanez puts it:

Given its many limitations, it is unlikely that AI will completely replace human translators. Indeed, in the case of technical, marketing, financial, legal, or medical content, the slightest error or ambiguity can have damaging consequences, generate disputes, and have a major impact on the company's reputation.


The Benefits of Using MT & AI in Translation

While machine translation and AI are unreliable when used by people who are unfamiliar with the languages or cultures of their text, they can actually be beneficial when used by a qualified and experienced translator or other language professional.


In today's globalised world, there is an increasing demand for translated content, and machine translation and AI can help translators meet that need.


Globally, people are consuming content at an increasingly rapid pace, and as a result, the demand for translated content has reached an all-time high. From blogs to e-books to on-demand TV shows and e-commerce websites, today’s companies must translate large quantities of content — quickly — to consumers.


MT and AI can increase the efficiency of translators, allowing them to get more work done in a shorter amount of time. In fact, translators have been using CAT (Computer-Assisted Translation) tools and TMs (Translation Memories) to improve their efficiency and ensure terminology consistency in their translations since the 1950s!


For those who are unaware, CAT tools and TMs don't "do the work for you", they merely segment the text out into chunks and remember how the translator decided to translate certain words or phrases. The tools can then suggest these translations when the same words or phrases come up again in the same text, giving the translator a reliable way to ensure consistency in their work.


While AI and machine translations are unreliable and unsuitable to be used in their raw form, they can be improved with human intervention. Post-editing of machine translation (commonly known as PEMT or MTPE) involves comparing a translation to the source text and making amendments to the MT output to bring the text up to publishable quality. The translator ensures that the translation is accurate and fixes any mistranslations, omissions, repetitions or strange additions to the text, as well as grammatical and punctuation errors. This is a job that only a professional translator fluent in both languages can perform well.



Conclusion

While machine translation and AI have made considerable progress in recent years, they still cannot match the quality of a translation provided by a professional translator.


In particular, languages that are grammatically and culturally very different from each other, such as Asian languages and Latin languages, are still nearly impossible for machine translation and AI to translate between.


Certain texts, however, should never be translated by a machine due to the importance of the text or creativity required. These include (but are not limited to) marketing, gaming, medical, patents, legal, financial and literary texts. These kinds of texts should always be translated by a professional human translator who specialises in the field.


However, machine translation and AI can still be useful in the translation industry, especially when post-edited or used by professional and experienced human translators.


Translators need to adapt and learn how to incorporate technology into their work flow in order to increase their efficiency and meet the ever-growing demands of our clients.


 

Sources:

"The Impact of AI on the Future of Translation"


"Machine Translation Evolves... But How Does it Impact Translators' Jobs?"


"The Future of Language Technology: The Future of Machine Translation"


"What are the Advantages and Disadvantages of Using Machine Translation?"


"Cat Tools: Unlocking the Potential of Computer-Assisted Translation for Global Growth" https://phrase.com/blog/posts/cat-tools/

Comentários


bottom of page