Whats the data weights going into the source? Seems a bit off to not clarify the relative linguistic contributions going in, to characterise them by source language coming out.

Given the amount of french poetry in the courtly love cycles, I'd be amazed if the corpus included them at high volume reporting such low french contributions. On the other hand if it's saga heavy, or has a lot of German romanticism influenced stuff like the sublime (good french word that...) then well.. DUH