The past few years have seen a massive growth and pushing forward in the public domain of artificial intelligence translation solutions which have revolutionised the world of professional translating.
- A word of warning!
- Suitability of AI translations for corporate clients – are they fit for purpose?
- Dumping prices at language service providers (LSPs) – breaking a lance for translators
- Realistic and commensurate charges
- 7 serious drawbacks of machine translations: a synopsis
- 7 serious drawbacks of machine translations: a full analysis
Practicable machine translation software has been available for around 20 years, whereby the quality frequently left much to be desired. Looking back at the results today, they were actually often rather awful.
Today’s solutions are based on Neuronal Machine Translation (NMT) engines and are capable of producing sensational results at a phenomenal speed (the observations stated here are based on translations from German into English, our core competent field).
And they are not actually that new. We have been using DeepL as a paid translation aid service (i.e. translated content is not stored on DeepL’s servers, which means that there are no data privacy issues) since 2018, and the improvement in the rendered translations has noticeably improved over the past 1-2 years.
What has changed, though, is that such tools have rapidly become omnipresent and have resulted in a radical change in the mindset of corporate clients. Many see in the technology an opportunity to cut costs in what is actually, as far as their budgets is concerned, a comparably insignificant cost driver. What they do not consider, unfortunately, are the potentially far-reaching consequences of transferring their reliability on professional translations to a blind faith in cheap and immediately available AI translations.
A word of warning!
Whether AI translations are the right answer for your company is subject entirely to your expectations and aspirations regarding your company’s public image and sales goals, whether you are willing to risk lost sales, damage to your company’s reputation or even its existence, or the loss of a potential business partner, or if you want to gamble on being passed over in favour of the competition following delicate contract negotiations.
The suitability of AI translations for your company’s documentation and whether they will be fit for purpose must be thoroughly and carefully evaluated. Content that is not mission-critical (such as quickly outdated email traffic or product reviews, etc.), or general content such as company histories or mission statements, can certainly be translated with AI with minimal revision without any notably detrimental results. Technical documentation, on the other hand, which invariably contains company-specific terminology, for instance, cannot be left to the inevitable vagaries of AI translations without inherent risk.
It all comes down to saving costs in completely the wrong area.
This is namely the crux of the problem.
A differentiation must initially be made between large-volume translations that are required faster than can be provided by a human translator, and the translations of machinery operating manuals, for instance, which will be used throughout the machinery’s service life und must be 100% accurate to prevent potentially disastrous and/or fatal consequences.
The costs of translating such operating manuals compared to the prices charged for the machinery itself are so negligible that it is incomprehensible why manufacturers seek to save a few euros by choosing the cheapest solution for their translations.
In a worst case scenario, the consequences can be disastrous.
Dumping prices at language service providers (LSPs) – taking a stand for translators
Another ramification of this translation revolution (from the translator’s point of view) is that the word or line prices being paid today to professional translators (whereby the emphasis here is on ‘professional’ i.e. ‘expert’) by translation agencies have consistently fallen, while the expectations to receive excellent quality have not. Understandably, major companies requiring large volumes of translations in multiple languages resort to large-scale translation companies for their needs.
Major translation agencies, however, have for years paid their freelance translators notoriously poor rates (and yes, this claim can be substantiated, but preferably not publicly), and the situation has deteriorated since the widespread predominance of AI translation solutions.
This is made undeniably evident by the prices that such companies offer their direct clients on their websites. A company that quotes prices of, say, 11 cents per word for a translation will obviously only pay its translation suppliers a proportion of this amount. This is far below the remuneration that, in our experience, was being paid 5 or 10 years ago.
Realistic and commensurate charges
A realistic end price charged by a translation agency for a technical translation in the machine engineering sector, for instance, should lie at around 18 cents per word. This would allow the agency to pay its freelancers (for they form the vastly overwhelming majority of an agency’s workforce) a commensurate remuneration of 11 or 12 cents per word.
An agency that offers its end clients a word price of 10 cents cannot afford to pay competent translators a reasonable rate. Consequently, they will not be able to engage proficient translators who are prepared to work for peanuts.
Corporate clients are quick to point to AI translation methods, thinking that these will significantly reduce the costs and processing time required for their translations. These claims are for the most part based on total ignorance about translating processes and everything they involve, and they also absolutely neglect the translator’s invaluable experience in, and knowledge of, the client’s specific field. Experience and knowledge that is commensurately otherwise rewarded as a matter of course in practically every sector.
7 serious drawbacks of machine translations: a synopsis
1. CAT (computer-aided translation) and glossary tools are absolutely imperative for corporate translations. Such tools cannot be reliably integrated in an AI translation workflow without human interaction.
2. NMT tools cannot recognise that a company may use specific terms which, although actually incorrect, they do not want to change to avoid confusing their customers and users (from our experience a true scenario!).
3. NMT tools cannot reliably handle the desktop publishing layout and formatting issues that frequently arise in the translating process.
4. Content for different sectors and industries requires a different approach, which cannot necessarily be reliably identified by an AI translation engine.
5. NMT tools have neither a brain nor a soul. They only regurgitate what they have been taught.
6. They cannot think out of the box to create imaginative translations.
7. They cannot identify errors in the source text and yet deliver the correct translations, nor can they notify the client about such cases.
7 serious drawbacks of machine translations: a full analysis
A summary
Corporate clients must continue to value the quality of the translations provided by human translators who legitimately use AI in their workflow, freeing up time for research and accuracy purposes to reliably deliver translations of the utmost quality.
Corporate clients must continue to appreciate the added value that premium quality translations bring to bear for their business success.
AI translation solutions are obviously here to stay.
They will inevitably over time quickly produce better and better results – but primarily only for specific purposes, however, and not necessarily in every corporate environment.
Translators must take this progress into account in the services they offer. Their remuneration (be it from translation agencies or end customers) must, however, be commensurate with their specialist knowledge, experience and technical skills (for example where translation software and its capabilities are concerned).
Translation agencies must learn that AI translation technology is not a means to an end in order to offer their clients rates which disparage the skills and experience of their freelance translators.
The suitability of AI translations for specific fields must be carefully considered by corporate clients.
AI translations may only be regarded as being the be-all and end-all solution for corporate translations at enormous risk.
Corporate clients must continue to value the quality of the translations provided by human translators who legitimately use AI in their workflow as it frees up time for research and accuracy purposes to reliably deliver translations of the utmost quality.
Corporate clients must continue to appreciate the added value that premium quality translations bring to bear for their business success.