-
Notifications
You must be signed in to change notification settings - Fork 1
Description
From Dominik:
For some datasets, it makes sense to label the voters. This is especially true when voters are criteria (such as city rankings https://www.preflib.org/dataset/00050) but also when the same voters vote across several elections (such as in the breakfast items dataset https://www.preflib.org/dataset/00035).
It's not clear how to implement this, since the standard format labels each ranking with its multiplicity (and implicitly assumes that it is okay to reorder rankings), so it's not just a straightforward addition of metadata fields.*
Initial Answer:
I think the easiest would be to add a metadata "voter labels", one per line (i.e., one per unique preference) that is a list of all the labels for the voters having this preferences.
Problem, if 10k voters have the same preferences, this list contains 10k elements.
Otherwise, we would need to completely change the format (somewhat leaning towards the pb format).
Do you have other idea Dominik?