slczgwh / REDN

Downstream Model Design of Pre-trained Language Model for Relation Extraction Task
MIT License
95 stars 14 forks source link

This is Relation Classification Not Relation Extraction #17

Closed sidney1994 closed 3 years ago

sidney1994 commented 4 years ago

This paper uses gold entity information rather than the predicted entities to evaluate the model. It is inappropriate to say that you achieved the best results on NYT and WebNLG, since previous works (such as HBT) did not use the gold entity information when testing.

slczgwh commented 3 years ago

See this.