Inferential Decompositions
Repository for Natural Language Decompositions of Implicit Content Enable Better Text Representations, accepted to EMNLP 2023.
Update 11/17/2023:
- We have released a notebook with a step-by-step process to generate inferential decompositions. Please refer to decomposition_tutorial.ipynb for more details.
- We have also released the text and inferential decompositions from the legislative tweets dataset used in Section 5 in our paper for analyzing legislator behaviour.
We're excited to hear what you use our method for! Please reach out with any questions or comments, or create an issue.
The COVID vaccine comment dataset is available upon request (all comments are publicly available, but we elected not to re-host data out of concern for commenters' privacy).
If you find our work useful, please cite us:
@inproceedings{hoyle-etal-2023-natural,
title = "Natural Language Decompositions of Implicit Content Enable Better Text Representations",
author = "Hoyle, Alexander and
Sarkar, Rupak and
Goel, Pranav and
Resnik, Philip",
booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
year = "2023",
publisher = "Association for Computational Linguistics",
url = "https://arxiv.org/abs/2305.14583",
}