Content Translaiton supports several machine translation services. Some of those services such as Apertium and MinT do not support HTML, and they only work with plain text. In these cases, rich text elements from the original content such as links and references get lost in the translation.

Content translation tries to re-apply those elements from the source paragraph to the translated version returned by the service. Since the translated text is different from the source, this mapping cannot be 100% correct.

This is an Epic ticket documenting in their sub-tasks different cases where the algorithm could be improved.