In the modern world of technical communication, companies face a wide range of challenges in their efforts to produce ever more extensive and complex documentation in increasingly shorter turnaround times and to make it available in a growing number of languages – all the while consistently maintaining the highest quality standards. At the same time, competitive pressures and cost pressures are also rising all the time. 

Here, the use of artificial intelligence (AI) in the translation process can open up new ways to save money without having to make any compromises in terms of quality – provided the process is correctly set up.

Combining AI and human expertise for an efficient translation workflow
Our tried and tested, multi-stage translation workflow combines the strengths of modern technologies with the skills and competence of experienced technical translators:

1. Pre-translation with Translation Memory Systems (TMS): Content that has already been translated and approved previously is taken from the customer-specific Translation Memory Database (TMS). This saves time and ensures that terminology is used consistently.
2. Translation of new texts with AI-assisted tools: AI-based machine translation is used for new or modified text segments. Today, tools such as DeepL or similar systems can already deliver surprisingly successful results in many scenarios – particularly for standardised technical texts.
3. Quality assurance through experienced, native technical translators: The machine-translated content is checked by qualified technical translators working in their mother tongue. Here, attention is paid not only to linguistic accuracy and correctness, but also to the use of required specialist terminology, consistent style and terminology, as well as to many other stylistic details.
4. Transfer back to the translation database: Once the translations have been approved, they are fed back into the Translation Memory Database – a continuous process of establishing knowledge and implementing quality that helps to further optimise future translations.
    

Quality begins with the source text 
The quality of the final translation is inextricably linked to the quality of the source text. Clearly structured, standardised and target group-appropriate source texts form the basis for successful machine pre-translation and fast, efficient post-editing.

Where AI works well – and where it struggles
AI tools are not a one-size-fits-all solution for translating all kinds of text. Their suitability depends to a very large degree on the text type, language pairing and context:

Very well suited:

• Operating instructions
• Data sheets
• Technical brochures

These types of text are often highly standardised, use consistent terminology and follow a clearly laid-out structure.

Less well suited:

• Software texts used in a UI context
• Marketing documents with a more creative
• style of language
• Legally sensitive content

Linguistic nuances, cultural differences and creative writing demand human expertise.
In addition, the translation quality also varies depending on the language direction. While AI systems tend to deliver solid results in languages such as English, French or Spanish, more care needs to be taken with less common languages.