23/01/2025

International software companies are under a lot of time pressure: every update must be localised quickly and efficiently for different markets. The translation effort is often enormous, especially for large manuals, user interfaces and support documentation. Using artificial intelligence in combination with professional post-editing can significantly reduce the time and costs involved.

Localising software, especially in multilingual environments, has become a crucial factor for success at many companies. Global competition requires software products to be adapted precisely and efficiently to different markets. This is precisely where the combination of machine translation and post-editing (MTPE for short) comes into play. This is not simply automated translation: it is a hybrid approach. The machine does the initial translation and experienced translators check the text and correct it as necessary. This approach is particularly advantageous for software localisation. This is because, in addition to high translation volumes, it demands technical accuracy and cultural adaptation. MTPE can do this and save a considerable amount of time and effort for large documents, such as manuals, without you having to compromise on quality.

While MTPE is ideal for continuous texts, such as software manuals or instructions, the situation is different when translating GUI (Graphical User Interface) texts. Texts in the software interface often contain little context, which is problematic for machine translation. This is because MT systems, such as DeepL, work based on word probabilities. The context within a sentence is the decisive factor. Short, contextless sentences or single words, as in software interfaces, provide unreliable results. This means that the end product may be either completely correct or completely wrong. This significantly increases the effort involved in post-editing and reduces or even eliminates the time and cost benefits that the company hoped to achieve.

GUI texts should therefore always be translated manually, but they can then be very helpful for MTPE as a glossary, which can be used to translate the associated manual, for example. This means that the system has access to the translation of the GUI and applies it correctly, which reduces the amount of post-processing required. An example from one of our software projects shows the savings made.

• 853 GUI entries were human-translated and integrated as a glossary. The GUI texts appeared approximately 4,000 times in the manual file. This means that there were 4,000 cases in which the terminology in the machine translation output might have had to be corrected.
• Without using the glossary, the machine translation would have produced 1,517 errors. Therefore, 38 per cent of the terminology occurrences would have been translated incorrectly by the MT system.
• After the glossary was integrated, there were only 568 errors by the MT system, i.e. only 14 per cent. This significantly reduced the costs involved.

A glossary therefore offers the possibility of making machine translation more efficient.
This is because the machine recognises specific or technically complex and sometimes cryptic terms using the glossary and translates them as specified. The human translator then takes over the fine-tuning, checks the machine translation output and corrects errors or checks for ambiguities.

One of the main reservations people have about machine translation is based on concerns about the quality of the translations. Precise terminology is particularly important for complex technical content. MTPE offers a clear advantage here: the machine provides a quick translation draft, while the human component ensures that subtleties and specific terms are used correctly. This step is essential in particular for software localisation, where misunderstandings can have serious consequences. Companies that use MTPE can thus ensure that their software meets the expectations of the target market from both a technical and a linguistic perspective.

But watch out: not every MT system is equally suitable for software texts. For the MTPE process to be successful, it is crucial that the selected MT system can handle labelled GUI texts. It is therefore important to compare the systems before implementing them and ensure that the chosen system meets the specific requirements of software localisation. Companies that take the time to find the right MT system can benefit significantly from the advantages of the MTPE process.

As with international marketing campaigns, cultural differences can also lead to stumbling blocks in software localisation. What has a positive effect in one country can be completely misunderstood in another.

An example from the marketing sector illustrates this. A US software provider used the slogan “We’ve got the guns” to emphasise the protection of its users against viruses. In the USA, this was a positive message that suggested power and strength. However, a literal translation into German would have had negative connotations. By adapting the translation to “Schutzschild” (protective shield), the company was able to successfully adapt its message to the target culture. This approach is also crucial for software localisation in order to convey culturally sensitive content correctly. This is another reason why the “human in the loop” approach is so crucial for MTPE. When it comes to cultural adaptation and transcreation of content and slogans, translation engines naturally reach their limit.

MTPE offers companies real added value in software localisation at a time when efficiency and speed are decisive factors for market success. By combining machine translation and human post-editing, time and costs can be saved without compromising on quality. Especially for companies that regularly need to provide software and documentation in several languages, MTPE is a forward-looking solution that combines flexibility, precision and efficiency and enables significantly shorter translation turnaround times. This is how companies manage to keep up with the times in a fast-moving world of regular releases and software updates.

Do you need software translations that can be used effectively and convincingly in any target market? Talk to us! Our localisation experts will advise you on the best process and on using MTPE. Our free checklist (only available in German) for software and website localisation helps you to assess where your processes are today and helps you to identify further potential for optimisation. Our case study shows how MTPE is accelerating translation cycles and optimising processes in practice with the software developer EPLAN.