26/03/2025

The EU Machinery Regulation 2023/1230 comes into force in 2027 and will present European companies with new challenges. Particularly in the area of technical editing, the workload increases considerably – and with it the costs for translation into other languages. Many companies are pinning their hopes on artificial intelligence (AI) and machine translation (MT) to cope with the additional workload. However, this approach involves risks, especially for sensitive technical documentation. We show how companies can make massive cost savings with existing translation technologies, and where the use of AI translation offers real added value.

The new EU Machinery Regulation significantly extends the scope of responsibility for manufacturers and distributors of machinery. The regulation will cover not only physical products, but also software and digital components with safety functions. In general, documentation requirements are increasing and in some cases even include descriptions of the source code and programming logic for complex machines. Manufacturers are obliged to create comprehensive technical documentation to demonstrate that their products also fulfil the safety requirements with regard to digital and software-based risks. However, the forthcoming regulation not only affects manufacturers, but also places obligations on all economic operators who place machinery and associated products on the market or put them into operation in the EU.

For many companies, the expanded requirements mean a considerable amount of additional work in the preparation of technical documentation. As documentation within the EU has to be translated into all the languages of the target markets, the cost of translation services increases in proportion to the growing volume of text. For companies that market their products in many EU countries, the financial pressure will therefore increase noticeably.

In view of the rising costs, the use of AI and MT might appear to be an obvious solution. They promise minimum human intervention and maximum cost savings. But the reality is more complex, especially when it comes to technical documentation, where accuracy and freedom from errors are top priorities.

Although machine translation systems deliver impressive results, they reach their limits when it comes to technical texts with a high degree of specialisation. The quality varies greatly depending on the language combination and subject area. While the German-English language pair often delivers good results, other language combinations may perform significantly worse. One reason for this is the different availability of training data and the fact that texts are often translated via English as a relay language in the background. If there are also a large number of translation project specifications to be observed, such as terminology or style guidelines, the quality of the purely machine translation output may be inadequate.

Another problem is the lack of reproducibility. Because MT systems develop continuously through regular training, the same source document can produce a different result for each attempt. This is a critical factor in the regular updating of technical documentation, which requires a high degree of consistency and accuracy. Added to this is the often inconsistent use of specific terminology, which can have fatal consequences, especially in safety-relevant texts.

DIN EN IEC/IEEE 82079-1 recommends the use of professional translators for instruction manuals for good reason – because the high level of technical knowledge required and the specific issues involved in translating technical documentation demand specialised expertise and a deep understanding of the subject matter.

Instead of relying purely on machine translation, the most efficient approach is to use an intelligent combination of technology and human expertise. There are three solutions that have proven themselves in translation practice, creating cost-efficient and high-quality translation processes:

Translation memories (TM) store translated texts in segments and make them available for future projects. TMs offer considerable savings potential, especially in technical documentation, which often contains similar wording and recurring passages. If a segment or sentence is repeated in a later translation job, the TMS provides the translator with the existing translation, so identical content only has to be translated once.

Translation memory systems are also extremely helpful for quality assurance. Intelligent features identify mistakes such as typing errors, transposed numbers or formatting errors. This can substantially increase translation productivity, while significantly improving the consistency of the content.

To get the best possible results from a translation memory, you need a solid foundation of data. But with every translation project, the TM grows – often uncontrollably. Duplicates, unclean segmentation and outdated terminology mean that translation memories cannot realise their full potential. In the worst case, identical matches are not recognised even though texts have already been translated, or inconsistencies arise due to different translation variants.

A regular language data clean-up not only improves the reuse rate and thus the savings on existing translations, but also reduces the effort involved in quality assurance. In addition to the translation memory, the terminology database should also be checked regularly. Contradictory entries, outdated technical terms or incomplete definitions have a direct impact on translation quality. A clean terminology database used as the basis for glossaries in MT systems offers particularly high potential savings – especially with regard to the technical language requirements of the Machinery Regulation.

A clean data foundation is essential, especially when AI applications or machine translation systems are used. AI systems learn from the data with which they are trained, incorporating any inconsistencies or errors in the data and, in the worst case, actually amplifying them. Targeted clean-up before training significantly improves the quality of machine translation output and avoids unnecessary costs due to incorrect output that requires extensive post-editing.

By using specialised analysis tools such as oneCleanup, the potential for cleaning up TMs and terminology databases can be quickly identified and implemented. Investing in clean language data pays off in several ways: through accelerated translation processes, improved quality and reduced costs – precisely the factors that businesses need to efficiently meet the extended requirements of the Machinery Regulation.

Leverage the savings potential and efficiency of AI translation while still meeting the high-quality standards of the Machinery Regulation. This balancing act can be achieved thanks to MTPE (Machine Translation + Post-Editing). The text is first machine-translated and then it is checked and optimised by professional translators. Depending on the text, it is possible to make savings of 10 to 30 per cent – without any reduction in quality.

The reality is that machine translation alone cannot fulfil stringent requirements. No AI or MT system is currently able to translate technical documentation or product descriptions into a target language accurately, completely and consistently. What is more, the sensitive information on source code, programming logic and safety functions required by the new Machinery Regulation requires a deep technical understanding. Choosing the correct terminology, using standardised formulations and accurately conveying safety-relevant information also require human expertise.

The MTPE workflow can be seamlessly integrated into existing translation processes. Combining it with translation memories and terminology databases creates a highly efficient process: existing translations are taken from the TM, new content is pre-translated by machine and the result is checked by subject matter experts. For the best possible results, the selection of the AI or MT system should be tailored to the domain and language combination. Incorporating company-specific terminology using glossaries can also significantly improve the quality of machine pre-translation and reduce the amount of post-editing required. With a professional MTPE service, businesses achieve a final result that is no different from a human translation – but can be produced significantly faster and more cost-effectively.

The Machinery Regulation is coming and will demand a considerable amount of documentation from companies. However, the associated translation workload can already be managed effectively through the combined use of technology and expertise. While translation memories and data clean-up increase efficiency and ensure consistency, MTPE enables cost-efficient processing of new content without compromising on quality. The human factor remains indispensable – as a guarantor of quality and for the review of safety-relevant content.

Instead of relying on supposed and error-prone cost savings through machine translation alone, companies should invest in sustainable translation processes. Careful maintenance of language data and the targeted use of translation technologies form the basis for long-term efficiency gains – and will help to meet the expanded requirements of the new Machinery Regulation in a cost-efficient manner.

Would you like to optimise your translation processes for the upcoming requirements of the Machinery Regulation? Our experts will be happy to advise you on the efficient use of translation memories, optimal tools for data clean-up and the use of MTPE. We’ll be happy to provide a no-obligation consultation.