Adding a masked-word supersense prediction task to BERT's pre-training, with WordNet supersenses as weak labels, improves performance on supersense disambiguation and Word in Context without human sense-annotated data.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SenseBERT: Driving Some Sense into BERT
Adding a masked-word supersense prediction task to BERT's pre-training, with WordNet supersenses as weak labels, improves performance on supersense disambiguation and Word in Context without human sense-annotated data.