@in2uitive will talk about his work on generating preconditions for events at @_starsem on Aug 6th 2:00 pm EST.
Joint work w/ @NateChambers.
๐งต๐for details.
Can we train models to o/p diverse preconditions for a given target event? The PeKo dataset can help:
https://t.co/whhSYy9bfa
But what to do when training has only 1 ref per i/p? See @in2uitive's @_starsem paper for an answer:
https://t.co/Aw4LtQvQto
6/6 We hope this dataset helps drive further research into event understanding and helps with schema learning, information extraction, and even story generation.
The dataset and preprint coming soon!
1/6 Glad to share our paper "Modeling Preconditions in Text with a Crowd-sourced Dataset" is accepted at #emnlpFindings@emnlp2020. We introduce a new large-scale precondition knowledge dataset (PeKo). Joint work with my great team: @b_niranjan, @NateChambers, and Mahnaz Koupaee.
5/6 We also frame a generation task -- given an event generate a precondition event and show how to derive training data using the PeKo annotations. Training on PeKo improves precondition generation compared to training over raw text or temporally ordered events.
Excited to share that our paper "Modeling Label Semantics for Predicting Emotional Reactions" has been accepted at #acl2020nlp ๐! Joint work with my amazing team - @b_niranjan, @NateChambers, @in2uitive and Mohaddeseh Bastan
Our paper on Multee: Repurposing entailment models for multi-hop QA is now on arxiv:
https://t.co/gjjcxatzTB
Work led by @harshjtrivedi94 w/ @in2uitive at @stonybrooknlp and @tusharkhot and Ashish Sabharwal at @allen_ai.
A short summary thread: