Even if Dorfladen is not ambigous, it could be disturbing (and at first reading be understood as some obscure compound of -fladen.
Yep - I agree with your perception. But the point was not {that use of ligatures vs not using them is for disambiguation} but instead {that only ambiguous compounding needs human intervention, whereas other compounding can (always?) be dealt with by the computer if a list of words to build them out of is available}. Any list containing "Dorf" and "Laden" can easily be used to avoid ligating these components. Perhaps this is what was meant, but I'm clarifying just in case.
Once I read a text, it used ligature (inappropriately) in the word Auflage 'obligation', which is compounded from the prefix auf- 'upon' -lage , a nominal derivative of 'to lay'. Anyway, it's one word with its own meaning. Because of that stupid ligature I read it twice as [ofla:ʒ], thinking it would be a yet-unknown French loanword, before finally realising it was simply Auflage.
You have a good eye, and I've had similar experiences. Interestingly your observation can be used as evidence that even well-lexicalized word[ usage]s can benefit from not being ligated at certain morpheme boundaries. Thankfully, lists are sufficient to address cases like "Auflage" as well. I'm saying it this way because I simply don't know whether there would be a similar effect if German didn't have French loanwords.
So we could be getting into the realm of AI-hard aesthetic judgments: Some cases that "should" ("Stickstoffflasche") don't actually depend on ligation, whereas others matter for other psycholinguistic reasons. Ideally publishers caring enough about ligatures have copyeditors/proofreaders paying attention to this.
Stephan
