Open Anaphory opened 5 years ago
Re: l5
, I would say that "explicit is better than implicit". If we know that a language is missing a feature, we should say so by putting -
in the data. If a value is not listed in the data, we should assume as little as possible and treat it as missing.
Re: l7
, what's a real world situation where you would code something like this?
Re:
l7
, what's a real world situation where you would code something like this?
“Our wordlist, which was cognate-coded by an expert on l7
, contains two forms for l7
. The expert said that the first form is in class A, but forgot to code the second form.”
There may be better use cases.
My synonym handling code implies a certain difference between
?
and-
. I would like to for various combinations of unknown, missing, absent, and present values, what the binarization and single-value summary that these should look like. Ideally, this would be in a format that could be turned into a list of test cases.F1: a ternary feature (ABC), to be binarized.
should in my opinion give the following alignments:
I'm not sure about
l5
, though. I think this is the one easier to implement, but a value not being listed in the data might equally mean that it is unknown.