Multilingual e-learning: courses in many languages

People learn best in the language they think in. Anyone working through a course in a foreign language spends part of their attention on the language itself and misses nuances. With compliance training this can be dangerous: a safety briefing that is not understood does not serve its purpose, even if it is formally completed.

Internationally active companies, universities with international programmes and organisations with staff from many backgrounds therefore benefit particularly from courses in several languages.

For a long time, multilingualism was a question of budget. Every language version meant a translation agency, voice recordings, new videos and separate maintenance with every change. Many organisations therefore offered only German and English and hoped it would do. Machine translation and synthetic voices have fundamentally changed this calculation: another language today is mainly a question of subject-matter review, no longer of production.

Multilingualism affects different parts of a learning platform, which are often handled separately.

If you want to offer several languages, plan for it from the start. Text on slides should be real text, not part of images, so that it can be translated. Graphics with labels should be created as vector graphics with text layers. And slides need spare room, because some languages need considerably more characters than German or English for the same content.

Also create a glossary of fixed terms: product names, department names, technical terms. Such a glossary helps machine translation as much as the people who review it, and ensures that the same term is translated the same way in every course.

Language models now translate quickly and to a high standard. For a lot of course content, a machine translation is enough as a basis. Limits show with technical terms, fixed company terminology, legal wording and exam questions, where even a small shift can change the correct answer.

A two-stage approach has proven itself. First the machine translates, then someone with language and subject knowledge checks the critical passages. It is important that this manual work is preserved: when the source text changes and is retranslated, the versions set by people must not be overwritten.

Narrated courses and videos need a separate audio track for each language. Synthetic voices make this possible in many languages without booking speakers. Captions complement narration and help people with hearing impairments, as well as everyone who learns without sound.

Producing many language versions costs computing time and money. One option is to pre-produce only the beginning of a course in each language and create the rest only when someone actually uses that language. That way costs arise only where there is demand.

Languages such as Arabic and Hebrew are written from right to left. A platform must mirror the interface: navigation, arrows, progress bars and text alignment. Slides with mixed text, such as Arabic body text with Latin technical terms, need particular care.

Localisation goes beyond language. Examples, names, currencies, date formats, legislation and cultural references must suit the target audience. A data protection course for employees in several countries should explain which rules apply across Europe and where national rules differ.

Universe offers interface, narration and captions in fifteen languages: German, English, French, Italian, Spanish, Polish, Swedish, Chinese, Ukrainian, Dutch, Russian, Japanese, Arabic, Hebrew and Hindi. For each language, authors choose “full”, “beginning” or “off”; in “beginning” mode the first slides are ready and the rest is produced while the learner watches. Lecturers can set course texts, quiz and exam questions in their own language; this manual work takes precedence over machine translation.