Yes, 
COMET by @ABosselut et al. Demo: https://mosaickg.apps.allenai.org Paper: https://arxiv.org/abs/1906.05317 It's not perfect, but it's better than anything out there. I dare say that my speculation is that one can't do better with entirely non-neural methods.
.@YejinChoinka is right about this. are there other models that do better?
(serious question, not a rhetorical question.)https://twitter.com/YejinChoinka/status/1188310808008130561 …
-
-
-
i don’t actually see how to directly apply comet to these, but am interested to hear more and was just reading about it:pic.twitter.com/23klv9X7Dl
- 2 more replies
New conversation -
-
-
@YejinChoinka has made a great challenge dataset (https://arxiv.org/pdf/1907.10641.pdf …) that is very hard for existing models (61.2% on a binary decision) but on which crowd workers do much better (90.2%). Note the really careful effort to avoid being misled by annotation artefacts. -
i have some issues with that specific test, which I may at some point clarify. i like the direction though.
End of conversation
New conversation -
-
-
"Commonsense Reasoning for Natural Language Understanding: A Survey of Benchmarks, Resources, and Approaches" https://arxiv.org/abs/1904.01172
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@GaryMarcus we have built a deep linguistic platform relying only on pure linguistics to approach understanding. Not common sense inference yet. Limited handling of pragmatics which will get better with time. More details separately.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.