Phase 3 CR / CR NR | Single Domain
Created on October 3|Last edited on October 13
Comment
Let's try and compare just Coref runs vs Coref + NER runs (for the coref metrics)
Main Experiment
Smaller Runs
NER embeddings into Coref
Just as it suggests. Nothing fancy.
Run set 2
6
Add a comment
Priyansh Trivedi •
Summary of NER Shared embeddings:
1. Overall it seems to have a negative effect
2. Looking at Onto P and Onto R is interesting since it seems that here the precision is always capped off and the difference lies in the recall. But this is weird because why would precision be capped off even in control runs?
Reply