Fixing exposure bias with imitation learning needs powerful oracles

Add code
Sep 17, 2021
Luca Hormann, Artem Sokolov
Figure 1 for Fixing exposure bias with imitation learning needs powerful oracles
Figure 2 for Fixing exposure bias with imitation learning needs powerful oracles
Figure 3 for Fixing exposure bias with imitation learning needs powerful oracles
Figure 4 for Fixing exposure bias with imitation learning needs powerful oracles

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon

We apply imitation learning (IL) to tackle the NMT exposure bias problem with error-correcting oracles, and evaluate an SMT lattice-based oracle which, despite its excellent performance in an unconstrained oracle translation task, turned out to be too pruned and idiosyncratic to serve as the oracle for IL.

View paper onarxiv icon

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon