A Comprehensive Guide to AI in Translation in 2024
We live in an increasingly connected world where overcoming cultural and social barriers continues to be a challenge when it comes to bridging the gaps between language and communication.
However, thanks to an increased use of artificial intelligence (AI) in recent years, the practical applications in translation have been remarkable.
In this piece, we’ll be taking a deep look into the various uses and applications for the translation industry. We’ll investigate why and how it works, and perhaps more importantly when it doesn’t.
Whether you’re a professional linguist, a student, someone looking to learn a new language, or a business looking at using AI for translation, this piece is perfect for you.
What is AI?
AI uses a series of natural language processing systems to understand the meaning of a piece of content to then produce similar work in another language.
This feeds into a large language model (LLM) which contains a huge multimodal dataset that becomes more and more accurate the more frequently it is used for translation purposes, allowing it to learn more about the cultural implications behind each translation.
At Anglia Translations Ltd, we have used a similar tool in Computer Assisted Translation, or Translation Memory as it is commonly known, for many years, and when used correctly alongside a language professional it can really help improve efficiency.
AI in Translation
No matter how smart an AI platform may become, it could never take the place of a human translator with experience and first hand knowledge of the languages they are translating, as well as their specialist knowledge of the subject matter.
Although AI has made some serious progress when it comes to the quality and coherence, it still struggles to comprehend creative requirements, cultural nuance, jargon and tone of voice.
The only way to ensure this is guaranteed is to bring a human being with the relevant experience into the mix to ensure messages are making the required impact in the target language.
There’s no doubt that AI can help to increase understanding between those with different mother tongues and help to speed up certain parts of the translation process, however the utmost care must be taken as AI still very much has its limitations – we’ll cover this in more depth later on.
So, with this being the case, how can you use AI for translation in the most effective and efficient way?
What Kind of Translation Tools Are Available?
The most common AI tools in translation, are:
Statistical Translation
Statistical translations are models able to understand considerable amounts of bilingual data to assess patterns and put together a knowledge base which forms connections between words and phrases in various languages to understand the relationship between them.
Although you’d seldom realise you were taking advantage of this technology, this process is used extensively by professional translators in a process known as machine learning translation and is also running in the background of your favourite language learning apps too.
Translation & Language Learning Apps
As mentioned, there are plenty of apps that are utilising advanced algorithms to translate your words in real time, such as the Google Assistant on your phone for instance. Often all you need to do is speak or write the words or phrases you wish to translate and you’ll receive the results in seconds.
While these aren’t always spot on when it comes to any cultural nuance, and will often appear overly formal, they are ideal for when you’re travelling or require a simple phrase translated in a common language.
Web Browser Translation Tools
You may have noticed that your web browser, such as Google Chrome or Safari, will now translate web content into your device’s default language.
These tools often rely on AI to detect the source language and offer instant translation for web pages, emails and other content too, although you may need to connect the dots in some cases as these are typically “like-for-like” translations.
This removes the need for copying and pasting any content into a translation tool and is a quick and easy way to get the jist of what is being said on the webpage.
However, this is not an acceptable way to translate your web pages as a permanent solution, as it is very prone to errors and mis-translations.
What Are the Real World Applications of AI in Translation?
Utilising AI for both professional and personal applications can play a crucial role in overcoming everyday language barriers and helping to bridge gaps between communities that may otherwise not have access to translation options.
Although, as we mentioned, human intervention is certainly required in translation, there are a huge range of applications where AI can be incredibly useful, and with the increased adoption of handheld technologies they can often be taken for granted:
Global Communication
AI has given us a fantastic opportunity to revolutionise how we communicate both personally and in business. Never before have we had access to another language at our fingertips.
It gives the opportunity to communicate while travelling, in business meetings or other international events. Even if you have limited knowledge of the language in question, you can open new pathways to personal and professional growth with the click of a button.
Cultural Understanding
AI translation has been able to give us access to culturally relevant information. Websites, books, education, healthcare literature and much more can be translated to gain a good understanding of the content within a matter of clicks.
This enables anyone in the world with a smartphone, tablet or computer to gain knowledge of a new language or culture that would otherwise have been unavailable to them without hours of study.
Tourism
Tourism has benefited considerably from AI platforms, particularly when it comes to travellers looking to learn a new language.
AI that produces real-time machine translations has made international travel easier and more accessible to a variety of tourists.
From reading signs and ordering from restaurant menus to communicating with locals and finding your way around, travellers from all walks of life are enjoying the benefits of having the confidence to communicate with communities all over the world.
International Exports
AI has streamlined and improved our ability to trade overseas. The likes of official documents, emails, exports, and other business literature can be translated in minutes. This has enabled an increase in business abroad, with statistics showing that international trade volume and trade has increased by 4-6% on average per year since 1995.
Medical Services & Care
AI in translation has also vastly improved access to crucial medical care options and services across the world. Organisations like Translators Without Borders are crucial in this regard, since they are able to aid in humanitarian crises by allowing healthcare professionals and patients to communicate with each other, to ensure they are getting the best quality care and crucial information about the environment and impending weather events too.
AI, chatbots, and language glossaries help them communicate in some of the least spoken languages in some of the most remote areas of the world too.
Video Translations & Subtitles
In recent years the interest of foreign films and TV has increased by 76% among Gen Z and Millennial viewers.
What you may not realise is that the subtitles you see on your favourite Netflix shows are produced by AI, which is able to transcribe and in some cases translate these shows for foreign audiences.
That said, there have been some examples around context and accuracy, most recently with the series Squid Game, where it was claimed that the subtitles were so bad that the original meaning was often lost for English-speaking viewers. So, while it does give us an unparalleled glimpse into new worlds and cultures, it is certainly worth remembering you may not be receiving the dialogue in the way it was intended.
Although we have probably all used AI in translation at some point in our lives, we should note here that it should only be used with the utmost care. Using AI in translation with no human intervention should only be used for assimilation – information only. If it were to be used for dissemination, to reach a wider audience, it is not reliable, and can lead to serious repercussions.
What Are the Applications of AI for Translation Professionals?
Auto Subtitling
With the use of AI, real-time subtitles can be utilised for videos and other visual media. Using rapidly developing voice recognition and natural language processors, you can transcribe and translate video dialogue and generate real-time subtitles in various languages.
This may be especially useful for those who are hard of hearing, or anyone who prefers to watch audiovisual content in another language, such as their mother tongue for instance.
Multimedia Content
Of course, translation in media isn’t just about dialogue. AI is also capable of translating other visuals, sounds and text onto the screen. This could be the text on road signs, mobile phones, noises heard off screen and other similar elements.
Utilising the strides made in visual algorithms and language processing, AI is able to pinpoint and produce these elements, to provide a viewing experience that may also aid deaf and partially sighted individuals too.
Document Translations
If you are required to translate a considerable amount of documentation, AI could be the ideal solution.
Allowing AI to do the bulk of the legwork with repetitive tasks means that you free yourself up to work on more complex, creative tasks.
AI platforms such as ChatGPT or Google Bard are certainly capable of helping you translate the bulk of the work in the most common languages, but it’s critically important that you quality check everything to ensure you can make any changes that reflect the culture and nuances in the target language.
AI translations are not really suited to high level strategic communications that are brand critical and user facing as this generally requires a more creative translation process.
Error & Correction
Although AI shouldn’t be used to create reams of in-depth copy, under the right parameters it can be particularly adept at detecting basic errors and correcting them in any translated content.
Very similar to the way that spellcheck can crawl through a document, various plugins, addons and other browser tools can be added to your desktop, PC or tablet to pick up even the smallest errors.
Translating Live Events
AI now means translators can enjoy unparalleled efficiency when it comes to simultaneous interpretation services for events, such as news conferences, summits, speeches and all manner of live streaming events.
Natural language processing and AI systems can be used to help the interpreter by generating the words spoken into multiple languages at the same time. However, once again we should certainly stress that this process does still need the expertise of an interpreter, AI should only be utilised to make their job that bit easier.
What Can’t AI Do for Translations?
We’ve been waxing lyrical about the virtues of the likes of ChatGPT and other AI tools, which have more than earned their stripes as a powerful tool for facilitating translations.
But, like any tool, AI has its boundaries, so it’s vital to understand not only where it can work well, but also where it struggles. This is absolutely crucial, as incorrect translation can not only be damaging, but sometimes fatal.
When to Avoid AI in Translation
When Looking for Cultural Nuance
While we tend to think of language as a collection of words and phrases, it really is far more important than that.
Any language is made up of historical and cultural importance that cannot be grasped by a machine. While something like ChatGPT is fantastic at interpreting the raw data of a request, it is unable to decipher the complex nuances that tie the thread between language and culture. AI does not ‘understand’ and it is not ‘intelligent’ in the same way as a human being which means that the output will only ever be as good as the prompt that it is given.
This is why it’s absolutely vital that a trained translator with vast experience of a language and its culture remains the most important focal point of a translation project.
Creative & Business-Focused Content
While AI can help to create a framework for a creative translation, that’s as far as it should go in this scenario.
Developing ideas, solving problems, persuading readers and writing emotional content all require input from highly skilled translation professionals as they require context, emotional intelligence and an understanding of human interactions in various cultures.
For those that have spent a long time creating their powerful marketing content to engage with their target audience, it would be careless to then trust AI to recreate that same emotion, tone and nuance in another language. AI is not reliable in this context and every output can contain critical errors that could be harmful to any business’ brand.
Specialist Translation
One of the most dangerous areas of AI translation is when you’re seeking to translate a rare or moribund language, legal documentations, business literature and financial information.
These certainly require specialist translators with in-depth knowledge of the subject material and all the languages they are working with.
Precision, accuracy and knowledge are all crucial, since any errors can have severe consequences, and AI certainly can’t be relied upon when it really matters.
What Languages Does AI Struggle With?
A recent study showed that tools like ChatGPT are not producing neutral content, or in other words, as the Founder recently admitted, AI does have “bias shortcomings”.
This has led to often sexist, racist and overall offensive translations.
This is often the case with languages which aren’t receiving linguistic input or increased language modelling. Training data has a massive bias towards English texts predominantly, followed by European languages. 85% of the data in Large Language Models is in European languages.
This means that for other languages, the output can be very poor. These include the likes of:
Mandarin Chinese
AI just isn’t compatible with Chinese languages at this time. Mandarin, in particular, is known as a tonal language with over 50,000 characters and an incredibly intricate writing system.
Arabic
Arabic is a language that relies on the intricacies of grammar and context, which creates quite a flexible language with multiple dialects. Most AI systems struggle with the fluidity of these languages, so much so that two new LLMs have been created specifically for the Arabic language.
Indigenous, Moribund & Rare Languages
This isn’t a surprise, given how heavily AI platforms rely on datasets and learning models. Languages with limited modern study pose a real stumbling block for almost all AI systems.
Mixed Script Languages
The likes of Thai, Japanese and Devanagari is especially tricky for AI. Many scripts are very sensitive, and even the minutest of errors can swing meaning and understanding considerably.
Takeaways
To reap the rewards of AI translation, whether you’re a professional or a keen linguist, it should only ever be seen as an accompaniment to human skills, knowledge and expertise.
It’s also important to remember that the responses you’ll get from AI platforms, such as ChatGPT and Bard, will only ever be as good as the prompts you’re able to give it. Knowing whether to offer more or less information and when is a skill that needs refinement over time.
There’s no doubt that AI will continue to advance beyond 2024 – many different forms have been in our lives for longer than we realise – but mainstream adoption will certainly continue, so there’s no need to shy away from it. We need to understand what AI is capable of, when to use it, how to use it and what its limitations are.
All outputs should be used mindfully with awareness of ethical implications and cultural bias and there should always be some human intervention involved. As Richard Baldwin, economist and professor, said, “you will not be replaced by AI, but by a worker who knows AI.”
At Anglia Translations Ltd, we never use AI as a standalone translation tool and never offer raw machine translation as a service, without our clients understanding the implications. All of our translators are highly experienced and qualified human translators who have expert knowledge of various languages and the complex cultures that surround them.
If you’d like to learn more about anything we’ve covered in today’s blog, or you’d like to speak to us in regards to any of our services, why not get in touch?
