https://github.com/bloomberg/kbir_keybart
Experimental code used in pre-training the KBIR and KeyBART models
https://github.com/bloomberg/kbir_keybart
Last synced: about 1 year ago
JSON representation
Experimental code used in pre-training the KBIR and KeyBART models
- Host: GitHub
- URL: https://github.com/bloomberg/kbir_keybart
- Owner: bloomberg
- License: apache-2.0
- Created: 2022-06-08T16:12:41.000Z (about 4 years ago)
- Default Branch: main
- Last Pushed: 2022-07-08T16:56:37.000Z (about 4 years ago)
- Last Synced: 2025-04-19T23:31:58.454Z (over 1 year ago)
- Language: Python
- Size: 29.3 KB
- Stars: 26
- Watchers: 5
- Forks: 3
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Description
This repository contains the experimental code used in pre-training the KBIR and KeyBART models as described in Learning Rich Representation for Keyphrases (https://arxiv.org/pdf/2112.08547.pdf) and to appear in Findings of NAACL 2022.
Some of the code builds on top of code from HuggingFace Transformers (https://github.com/huggingface/transformers) and also takes inspiration from SpanBERT (https://github.com/facebookresearch/SpanBERT)
# Running the pre-training
Use the two bash scripts for running pre-training for KBIR and KeyBART respectively.
# Accessing Pre-trained models
Models are uploaded to HuggingFace along with Model Cards describing usage.
KBIR: https://huggingface.co/bloomberg/KBIR
KeyBART: https://huggingface.co/bloomberg/KeyBART
## Citation
```
@article{kulkarni2021kbirkeybart,
title={Learning Rich Representation of Keyphrases from Text},
author={Mayank Kulkarni and Debanjan Mahata and Ravneet Arora and Rajarshi Bhowmik},
journal={arXiv preprint arXiv:2112.08547},
year={2021}
}
```
## License
KBIR and KeyBART are Apache 2.0. The license applies to the pre-trained models as well.
# Contact
For any questions reach out to mkulkarni24@bloomberg.net