I have some ideas for facilitating the creation of the file for whole word relations, but that's for the future.
From the currently certain aspects of the file's design:
automatically creating relevant relations covered by the morpheme based annotation (anything sharing the same ID in one column)
annotating relations only containing sound change (i.e. sharing the same gloss but being of different doculects) as such
if word a and b have a relation, and word b and c have a relation, then possibly also the relation between a and c should be annotated and could be derived automatically - but really unsure about that one, we first need to look at actual data
I have some ideas for facilitating the creation of the file for whole word relations, but that's for the future.
From the currently certain aspects of the file's design: