We discussed marking them in linguistInfo, but I am not sure how or whether these should be the case.
My assumption is that the FST should add some annotation in the analysis. The exclamation mark is not recognized by the relaxed analyzer. But removing the exclamation mark does allow the analysis which tends to add a +Voc in most cases (or a +Interj).
Should we just then remove the exclamation mark and rely on the analysis/FST to properly mark the analysis when generating the importjson?