Make homonym numbers optional? about omorfi HOT 3 CLOSED

This is a good suggestion. The homonym numbers are useful for generation too, for the cases where the POS is same but the inflectional paradigm isn't, e.g. laki : lain (law) vs. laki : laen (peak). In retrospect, that is probably the only useful homonym numbering, though changing to it will be a major overhaul.

A quick workaround to optionalise the numbers should be easy to concoct, I'll try if a simple replace rule will work for most...

In future I'd probably add a less strning-representation dependent generation api in any case.

from omorfi.

This is my very sophisticated workaround branch as a stopgap for my own use: frankier@b38669d

It seems even for Scandic loan laki and real original Finnic laki which are quite clearly distinct, some applications might know which one they have and be able to somehow retrieve the homonym number, while others simply can not or prefer to get both generations for some other reason. On the other hand, if there was a known cap on the number of homonyms, it could also be fixed up on the Python API side by looping through and trying all if this is preferable for some reason (transducer size?).

from omorfi.

Interesting for the vowel harmony homonyms. I'll keep it in mind.

from omorfi.