Subword models struggle with word learning, but surprisal hides it.

Bastian Bunzeck, Sina Zarrie

Browse the full ACL paper archive.