search

budzianowski / multiwoz

Source code for end-to-end dialogue model from the MultiWOZ paper (Budzianowski et al. 2018, EMNLP)
MIT License
867 stars 199 forks source link

Add results for DAIR #120

Closed tianjianh closed 1 year ago

tianjianh commented 1 year ago

I am the author of Robustness through Data Augmentation Loss Consistency. Huang et al. TMLR 2022.

I would like to add our results for DAIR, which is a consistency regularizer working at loss level to improve robustness of model. We empirically show that DAIR not only improves the performance with the presence of named entity replacements but boosts the performance under the standard test set. It worth nothing but the simple regularizer provides non-trivial performance increases.

Please let me know if you need anything for the approval of the pull request. Thank you!

jianguoz commented 1 year ago

Hi @tianjianh, thank you for your valuable contributions! I noticed that you've shared the code for your work, which will undoubtedly enhance its impact. Could you please provide the link to your code in your commit? Thanks!