The automated procedure surely has its advantages, but it is dangerous to use it without checking the data, and in this case, this becomes very clear. Since the senses offered by the author are so numerous, finding a hit is trivial, but this hit may often be misleading. So I would suggest to raise the bars and think of a future workflow by which we can avoid that you will end up comparing apples with pears, due to problematic mappings, since Concepticon was designed to make concept comparisons more exact in linguistics.
@LinguList I am trying to add custom columns "Year", "Etymology", and "Loan" to the EntryTable, but nothing is happening. I guess I have to define those new columns somehow in the beginning with attr.ib but I'm not sure how to do that, or if it's possible at all to do something like this.
I would suggest to tie versions to actual publications, or if you prepare one. If you make a version for every smaller change, Zenodo has to work a lot. Our procedure is: we do new versions always when we have an upcoming publication where we want everything to be clearly identifiable with this particular version.
@LinguList added this script to create backwards reconstructions based on previously extracted soundchanges, see column "rc" in forms.csv. Should turn this into a command-line command at one point.
Btw, the recostructions are customisable through the parameters, now I made it so that there are max 10 false positives per word. It is also possible to spell out the combinations and run some filters over it :)