In the current version, the bib creation process is stateless and does not resolve duplicate keys. It should resolve duplicates by adding a -b, -c, etc.
Relevant code: https://github.com/ir-anthology/ir-anthology-data/blob/master/scripts/util/rename.py
Keys are therefore also not stable across builds if sources change. Relying on the NTLK stopword list also means we have the risk of hidden changes that change key generation.
In the current version, the bib creation process is stateless and does not resolve duplicate keys. It should resolve duplicates by adding a
-b,-c, etc.Relevant code: https://github.com/ir-anthology/ir-anthology-data/blob/master/scripts/util/rename.py
Keys are therefore also not stable across builds if sources change. Relying on the NTLK stopword list also means we have the risk of hidden changes that change key generation.