Translation is a field where modern technologies based on artificial intelligence (AI) began to be widely used relatively early. Even early machine translation engines, based on statistical methods, are examples of using AI to translate text. However, there has always been palpable resistance in the translation community to automated translations. This is hardly surprising – machines could theoretically replace human translators and put professionals out of a job.
So, does AI-assisted translation bring more opportunities or threats? This question can be answered from two perspectives: the client who needs translation and the professional translation service provider, such as a translation agency or individual translator. It might seem that clients see only benefits, while translators face risks. But is this really the case? This article aims to explore this question.
AI Translation: How the Translation Industry is Using Artificial Intelligence
When discussing AI in translation, the first thing that comes to mind is probably machine translation and popular translation engines such as Google Translate or DeepL. Indeed, this is the primary way artificial intelligence is being used for translation.
Machine Translation: NMT and LLM
The last decade has brought a revolution in automatic translation, with neural machine translation (NMT) displacing statistical methods. NMT, based on deep learning technology using artificial neural networks, is much more accurate and better mimics translations made by human translators. As a result, it has begun to be widely used by professionals as well.
In recent years, competition for NMT engines has emerged in the form of large language models (LLM), such as ChatGPT. LLMs can be classified as a type of NMT since both approaches use artificial neural network technology. The main difference is that classic NMT engines are trained on bilingual corpora with the sole purpose of translating text, while LLMs are trained on much larger datasets and perform a range of natural language processing tasks, including translation.
Both approaches are currently used by amateur and professional translators alike. However, professionals use machine translation differently from ordinary users. At a translation agency, the process doesn’t involve simply dropping the text into an online machine translation application and then copying the translation to send to the client.
Professional translators and language service providers use CAT tools (computer-aided translation), integrated with paid versions of machine translation tools. These tools rely primarily on translation memories created from pre-processed texts and terminology databases—pre-prepared glossaries. They divide a text into segments and propose translations into the selected language. Thanks to integration with NMT engines, text segments not previously in memory are translated automatically. Each text segment, whether inserted from translation memory or machine-translated, is then verified by a human translator.
Artificial intelligence is also used in CAT programs to help proofread text, detect errors, and identify inconsistencies in translation. AI is used to create terminology databases based on translations previously verified by human translators.
Advantages of Using AI for Text Translation
Speed and Efficiency
One of the biggest advantages of AI translations is speed. Traditional human translations can take anywhere from a few hours to several days, depending on the text’s length and complexity and the translator’s skill and availability. AI-based tools can translate long documents in seconds. If the machine engine handles the text well and an experienced post-editor efficiently edits the text, the client can receive it faster than if a human had done it from the beginning.
Efficiency and Costs
Using machine translation can help reduce the cost of translation. While free automatic translation tools like DeepL or Google Translate have limitations regarding the amount of text translated and data protection, companies can invest in paid translation engines. This usually costs less than hiring professional translators, but the risk of errors and the need for human verification and correction of machine translation must be considered.
From a translation agency’s perspective, using machine translation with post-editing by a human translator increases work efficiency and allows offering a lower price to clients.
Disadvantages and Limitations of AI Translation
Uneven Quality and Errors in Translation
Both points mentioned earlier should be considered with an important caveat. AI can speed up translation and reduce costs only if it generates a correct, high-quality translation. However, artificial intelligence currently struggles to handle translation efficiently. As a translation agency utilizing AI technology, we observe that only a small portion of NMT-generated content can be used without corrections. Many AI-generated texts require significant translation intervention, and some are entirely unusable, making it faster to translate them “from scratch.”
Therefore, using machine translations without human verification involves a substantial risk of errors. If you opt for this type of translation, always ensure it is verified. The best results are achieved when post-editing is performed by a professional translator or a native speaker familiar with the subject matter.
Limited Scope: Languages and Specializations
The quality of AI translation heavily depends on the source and target languages. The effectiveness of translations, whether generated by classic NMT or LLM, relies on the content they were trained on. The largest datasets available for natural language processing algorithms are primarily in English, making it the language AI handles best. For less common languages, such as Polish or Japanese, machine translation results are significantly poorer and require more corrections.
When using machine translation, it is crucial to consider the type of text involved. AI performs much better with general, conventional product descriptions on websites than with specialized technical, legal, or medical texts. Furthermore, mistakes in translating content on an e-commerce platform are less critical than errors in translating contracts, technical specifications, or drug descriptions.
The Future of AI Translation
AI-based translation technologies offer opportunities for those who use them wisely, whether from the perspective of the translation service provider or the client needing the translation.
For large companies, AI-based translation technologies might be more advantageous to purchase and implement in-house than using external language service providers. This carries risks—if the implementation is incompetent, translation quality may be unsatisfactory. Proper implementation requires significant initial effort but can be financially and time-efficient in the long run.
Professional translators will need to use technology to make their work easier and faster, focusing more on post-editing translations than translating text from scratch. They should specialize and hone their language skills, particularly in less common languages, to operate in niches where machine translation is less effective.
For translation agencies like Bireta Professional Translations, keeping up with and adapting to the latest technological trends is essential. At Bireta, we began implementing AI-based solutions several years ago to offer our clients the best possible service, combining modern technology with the professionalism and experience of our translators and project managers.
One thing is certain: we will not escape artificial intelligence, and its use will become more widespread. Almost every translation process will soon start with AI tool processing. However, the work must not end there. The best results are achieved by combining AI tools with human supervision. Without skilled human management, achieving satisfactory results in text translation is impossible. Only companies that understand this will fully benefit from artificial intelligence’s potential.